Home

Relation between fine structures in hearing thresholds and distortion

1. APPENDIX D THRESHOLD AND DPOAE FINE STRUCTURE COMPARISON OF RESULTS o D o zd o E o if o N 52 vt i T e 52 vt Ca FE o oO o 2 2 C 2 d Nn Y O m e E d o 3 B B S lt ge e o f E o o o o o o ca 7 o o o o o o 8 7 T lt m w T Gee UN Pas gp epnijdure 3voda fds gp apnujdwe yoda A Das gp epnyjdwe pjouseiu L 1dS gp epnydwe pjouseuu L Ee Mds gp apnyidwe pjouseju L 1dS gp epnydwe pjouseuu L l a N e vt l e m N eo T e l e T EN e t l e T u m t T S m S 9 S LO LO q ive N N un e 9 L NT E d NT ro LA ue N m A a L3 u Q H H E S S B S a i o o o o o e TO o o o o o B o o o o o o B o o o o o o T T Wm T o m e Y T To WW YN TO PT Das ap epnyjdwe 3voda Das gp epnyjdwe 3voda fds ap epnyjdwe yoda Kas gp spmdwe 3yOda Threshold and DPOAE fine structure of subject 8 Figure D 8 94 Hds gp epnidure pjouseiu o o i o m N e vr 2500 o o o o 10 1 2 3 40 Das gp epnudwe 3voda Mds gp epnidure pjouseiu L e o qe e vr Mas gp epnyjdwe pjouseuu o o o o T o o o Q N T T T T 2500 at ft E 1600
2. o N en o o o e e e fds gp epnijdure 3voda Ss o o o o o ji I I fds gp epni dure 3voda 40 Threshold and DPOAE fine structure of subject 9 Figure D 9 Mas gp epnijdure pjoysasyL o o o o ce 1 o m N e vt 1600 oO o o O O coc E Tog ROT Dds ap epnyjdwe 3voda 1dS gp epnyjdwe pjouseuu e Ld o o o Q o N a Mas ap epnyjdwe pjouseiu o o o Q o T N e ar 2500 1600 att E E o o o e e fds gp apnyjdwe 3yOda o o o e o o N e Y I I i fds gp apnudue 3voda 2500 1600 1 2 ff f Threshold and DPOAE fine structure of subject 10 Figure D 10 95 1ds gp epnyidwe pjouseuu L 1dS gp epnyidwe pjouseuu L e e q o o o o o o e e o T N e o T N e UH E 2500 2500 E 4 1600 1600 Threshold and DPOAE fine structure of subject 11 APPENDIX D THRESHOLD AND DPOAE FINE STRUCTURE COMPARISON OF RESULTS o o o o o o e o o o o o E 8 9 7 T m moo coge DOES SN fds gp epniidure 3voda Pas gp epniidure 3yOda Das gp epnyjdwe pjouseiu L Das gp epnyjdwe pjoysesy 1dS gp apnyduwe pjoysesy Das gp epnyjdwe pjouseiu o N e
3. NE B E IN DPOAE as function of f Ripple height dB SPL o N Hu D 4 5 6 7 Subjects TM rti M 8 9 10 11 12 Average Figure 11 8 Ripple height in dB SPL from each subject The average threshold and DPOAE ripple height is shown in Figure as a function of 2f fo and f The height of the ripples is calculated in 1 8 octave bands and averaged over the 12 subjects The threshold ripple height fluctuates around 4 8 dB SPL on average with a standard deviation of 1 4 dB SPL The height of the DPOAE ripples is 8 dB SPL on average among the analyzed 1 8 octave frequency bands with a standard deviation of 3 9 dB SPL A specific trend as a function of frequency can not be found for the ripple height neither for the threshold nor for the DPOAE fine structure However the threshold ripple height appears to be less fluctuating than the height of the DPOAE ripples 11 3 Relationship between DPOAE and threshold fine structure Ripple height Ripple height 20 r r r r 20 r r r r 15 15 T T n n 10 0 10 a co a S 5 5 0 i i i i i i 0 i i i i i 1000 1250 1600 2000 2500 3150 4000 5000 1000 1250 1600 2000 2500 3150 4000 5000 1 4 Hz Hz Figure 11 9 Ripple height averaged across subjects in dB SPL for the threshold blue and DPOAE black fine structure analyzed in 1 8 octave bands The errorbars represent the stand
4. Subjects 1 2 3 11 Al Average dB SPL 32 10 06 0 5 1 2 Standard deviation dB SPL 1 5 1 1 0 4 0 5 1 4 Table 11 5 Height deviation in dB SPL of the coincident ripples in the test retest results height variates on average 1 2 dB SPL with a standard deviation of 1 4 dB SPL The high correlation between the test and retest results indicates that the implemented high resolution audiometer is reliable Therefore it can be used to track hearing threshold and proceed to the analysis of the relationship between threshold and DPOAE fine structure 11 3 Relationship between DPOAE and threshold fine structure A possible relationship between the threshold and DPOAE fine structure obtained in the experimental results is analyzed here First a study of the fine structure parameters ripple prevalence ripple spac ing and ripple height is made taking into account both the pooled and the individual results Then the subjects are gathered into different groups regarding their DPOAE measurement results The threshold and DPOAE characteristics shown within each group are studied in detail 51 CHAPTER 11 RESULTS An rough analysis of this ripple parameters is also presented in Appendix D where the threshold measurements are compared to the DPOAE measure ments as a function of the distortion product frequency 2 f1 f the primary frequency f2 and the geometric mean of the two primaries y f fg DPOAE p
5. co O CO Q gt CO E Og 7 3 fds gp epnijdure 3voda t o o o e o N e l I fds gp epniidure 3voda o n 1 2 f Threshold and DPOAE fine structure of subject 1 Figure D 1 1dS gp epnyjdwe pjouseu o T o a e 2500 1600 o o o o o o y T N e Y fds ap epnyidwe 3voda 1dS gp epnyidwe pjouseiu oO o o o Q o N e Das gp epnyjdwe pjoysasyL l o o Kc E 7 2500 1600 aff T o o T N E LH 2500 o e Ut 1600 o o o e e fds gp epniidure 3yOda pos o o oO x o8 i fds gp epniidure 3voda 30 1 1 2 f 5 Threshold and DPOAE fine structure of subject 2 Figure D 2 91 APPENDIX D THRESHOLD AND DPOAE FINE STRUCTURE COMPARISON OF RESULTS Das gp epnyjdwe pjouseiu L o o o o o T N a O coc O a E NC i fds gp epniidure 3voda Das gp epnyjdwe pjouseiu L o T o e o i o N a o 2500 1600 40 Das gp epnyjdwe pjoysesy o o o o o o Si e i Mds gp epniidure 3voda I Mds gp epnidur Threshold and DPOAE fine structure of subject 3 Figure D 3
6. 150 00 1250 1600 2000 2500 3150 4000 5000 2f f Hz Ripple spacing octaves 0 15 0 1 0 05 0 n i i i n i 1000 1250 1600 2000 2500 3150 4000 5000 Hz Ripple spacing Hz 350 300 250 200 00 1250 1600 2000 2500 3150 4000 5000 Hz Figure 11 7 Ripple spacing averaged across subjects in octaves up and in Hz bottom of the threshold blue and DPOAE black fine structure analyzed in 1 8 octave bands The errorbars represent the standard deviation between subjects function of f2 when compared to the results presented by Reuter and Ham mersh i Reut 06 Moreover this pseudo linear growth curve appears to be in concordance with the threshold ripple spacing as a function of frequency When looking at the results in octaves the average spacing curves present a flat tendency revealing a more or less constant periodicity Therefore the similarity on the threshold and DPOAE fine structure suggested in Chapter appears to be confirmed by the experimental results 11 3 3 Height The average and standard deviation of the threshold and DPOAE ripple height calculated for each subject within the frequency range from 1400 Hz 55 CHAPTER 11 RESULTS to 2800 Hz is presented in Figure The DPOAE ripples are in general more pronounced than the threshold ripples 20 Threshold 18 sz ips pies p I E E EH DPOAE as function of f 16
7. 1dS gp apnyidwe pjouseiu e E o o a vt E o ive N v e o oO e o o o dl 1 Se T fds gp apnujdwe yoda Mas gp epnyjdwe pjouseiu Das gp epnyjdwe pjouseiu L o o o o e C l e a e t e fo o o LO LO e e T N e o o o oO oO o e o o o o gt i dS ap epnyjdwe 3voaa i Has gp epnyydus Threshold and DPOAE fine structure of subject 4 Figure D 4 92 Dus ap a ploysery L o o o Se E eS LO e LE a o Hm o T Das gpl NNI avoda pes ap ji le piouseuu L Das gp epnyjdwe pjouseiu L o 2 o E o e o o ive ive N N a N T A amp i8 He o o m o e Y Ub Ts gp idu avoda 1 i fds gp epnjdwe 3voda Threshold and DPOAE fine structure of subject 5 Figure D 5 Mas gp epnijdure pjouseuu 1 o 1600 2500 f o o o o o fds ap epnyjdwe yoda 1dS gp epnyjdwe pjoysesy o Mas ap epnyjdwe pjouseiu O o N e n d 10 0 0 0 0 Has gp apnyjdwe 3yOda 2500 1600 o o o e o o fds gp epniidure 3vOda 1 2 6 1 Threshold and DPOAE fine structure of subject 6 Figure D 6 93 Mas gp epnyidwe pjouseuu L Mds gp epnyidwe pjouseuu 1 e e
8. The high correlation between the shape of the thresholds obtained by the two meth ods proved the validity of the FINESS algorithm However this comparison with the 3 AFC method revealed the limitation of the algorithm regarding accuracy on the absolute threshold level 7 1 FINESS procedure The subjects are instructed to press a button for as long as they can hear a tone The level of the stimulus decreases while the button is pressed and it increases when it is released The Matlab files necessary to run the audiometer are included in the enclosed CD The main function Audiometry Gui m constitutes the interface for the experimenter to proceed to the threshold measurements 28 7 1 FINESS procedure The stimuli consist of 250 ms tones with a stable interval of 200 ms and a raised cosine raise and fall of 25 ms to make the detection easier There are no silence intervals between the tones presented The level of presentation is adjusted to a dB HL scale according to ISO 389 and the calibration of the system see Appendix E The level of the tones varies with an attenuation rate of 3 dB s and the level is changed by 0 75 dB steps The frequency resolution is of 100 points octave The frequency points are screened in ascending order The small frequency steps size allows the possibility to have just one reversal per frequency The frequency steps are half the size of the minimum expected spacing between fine structure
9. dible range The area between 100 Hz and 7 kHz contains all frequencies guaranteeing a full speech intelligibility For this reason an audiometry is typically performed within this range Cran 08 A threshold of pain mal ATT 100 audible range 80 60 40 Sound Pressure Level dB 20r absolute hearing threshold 1 1 1 1 1 1 1 1 1 1 20 50 100 200 500 1k 2k 5k 10k 20k Frequency Hz Figure 3 1 The audible range of normal hearing humans between the ab solute hearing threshold and the threshold of pain Redrawn from and Soun 07 The absolute hearing threshold is the most examined limit of the audible range It is different for each person but shows a somewhat common shape for healthy ears As shown in Figure 3 1 normal hearing subjects are most sensitive in the range between 2 and 5 kHz whereas the threshold increases towards the boundaries of the hearing range Hence a relatively high sound pressure level is required in order to make sounds of low or high frequencies audible Since the threshold is strongly dependent on the frequency a reference has been developed according to numerous measurements on young adults free of all signs of hearing disease This reference is used to evaluate individual hearing loss compared to the human average Cran 08 There are different threshold references regarding the way of presenting the signal A number of psychoacoustic researches refer to a bin
10. subject 11 LEE 1600 2000 2500 Frequency Hz Figure B 4 High resolution results from subjects 7 to 12 dB SPL dB SPL dB SPL Threshold fine structure subject 8 1600 2000 Frequency Hz 2500 Threshold fine structure subject 10 Eu WEE 1600 2000 Frequency Hz 2500 Threshold fine structure subject 12 1600 2000 Frequency Hz 2500 DPOAE fine structure measurements Tests Journal The DPOAE measurements were performed in order to screen the DPOAE fine structure of the subjects The ILO96 Research system from Otodynam ics was used for this measurements The DP gram test of the ILO s DPOAE function was configured to fulfill the requirements of the system specified in Section C 1 Setup The DPOAE measurements test was performed in the Audiometry Room B4 103 at Aalborg University Fredrik Bajers Vej 7 9220 Aalborg E The system s setup is shown in Figure ILO96 Sofware Amplifier unit Ear probe Figure C 1 Setup used for DPOAE measurements test The test subjects were seated in the Audiometry Room where the equipment was placed The ear probe had to be properly fitted into their ears 84 C 2 Equipment C 2 Equipment The equipment used in the DPOAE measurements setup is detailed
11. 4000 Figure C 2 Distortion product otoacoustic emissions and background noise measured and DPOAE ripples detected in subjects number 1 to number 6 as a function of fa 88 C 4 Results dB SPL dB SPL dB SPL 15 10 a o 15 30 Subject 7 T LU EA 1600 2000 2500 f Subject 9 3150 4000 Il 1600 2000 2500 f Subject 11 3150 1600 2000 2500 f 3150 4000 dB SPL dB SPL dB SPL Subject 8 5 Halll 1600 2000 2500 f Subject 10 3150 B HEIC TEILE 30 i al i i si 1600 2000 2500 3150 4000 f Subject 12 15 10 H 1600 2000 2500 f 3150 Th 4000 Figure C 3 Distortion product otoacoustic emissions and background noise measured and DPOAE ripples detected in subjects number 7 to number 12 as a function of fo Threshold and DPOAE fine structure comparison of results A comparison of the results obtained from the threshold screening and the DPOAE measurements is roughly presented in this appendix The figures presented in the following provide a v
12. According to the data obtained from the cali bration this would result in a sound pressure level of maximum 8 8 dB The sound pressure level and the hearing level according to ISO 389 is presented in Figure 104 E 2 Calibration Data 0 9 4 0 8 H 4 0 7 4 Electric Noise uV 50 100 200 400 1k 2k 4k 10k Frequency Hz Figure E 4 The electric noise measured at the terminals of the head phones 10 ash J hoe 20 ig J 25b 30 H E Level dB 35 E 40 H E 50 50 100 200 400 1k 2k 4k 10k Frequency Hz Figure E 5 The sound pressure level light gray and the hearing level dark gray produced in the headphone as a result of the elec tric noise 105 APPENDIX E HIGH RESOLUTION AUDIOMETRY CALIBRATION AND NOISE ESTIMATION E 3 Application of the Calibration Data The data from the calibrations measurements is used to ensure an exact presentation of the desired hearing level with the audiometric equipment specified in Appendix B The reference equivalent threshold sound pressure levels for Sennheiser HDA 200 circumaural headphones are given in ISO 389 8 2004 E These values are used to calculate the corresponding sound pressure level to a given hear ing level Due to the low resolution of the given values interpolations are necessary From Section E 2 1 following relation bet
13. CHAPTER 1 INTRODUCTION AND PROBLEM STATEMENT the frequency resolution leads to a longer duration of the experiment which influences the subject s concentration An audiometric algorithm to detect fine structure in the hearing threshold has to be developed It has been chosen to use the FINESS FINE Structure Screening algorithm developed by Heise et al Heis 08 which detects threshold changes using a fixed frequency tracking procedure based on the variation of the presented levels The DPOAE measurement will be performed with the commercial ILO96 research system using a high frequency resolution In order to analyze the different components involved in the DPOAE generation a wider frequency range will be measured compared to the high resolution audiometry The results of both methods the high resolution audiometry and the DPOAE measurement will be analyzed in the same frequency range in order to find similarities and differences The analysis is extended by the use of cochlear models to study a relationship between the fine structure of the hearing threshold and that of the DPOAE The human auditory system A basic description of the anatomy and functionality of the human auditory system is given in this chapter to support further explanation of the nature of hearing threshold and OAE fine structure The human auditory system can be divided into three parts the outer ear the middle ear and the inner ear The different ele
14. Hammershgi Distortion product otoacous tic emission of symphony orchestra musicians before and after rehearsal Journal of the Acoustical Society of America Vol 1 No 121 January 2007 S M Ross Introduction to Probability and statistics for Engi neers and Scientists Elsevier Academic Press 3rd Ed 2004 L A Shaffer R H Withnell S Dhar D J Lilly S S Good man and K M Harmon Sources and Mechanisms of DPOAE Generation Implications for the Prediction of Auditory Sensi tivity EAR amp HEARING October 2003 Sound amp Hearing Widex 3rd Ed 2007 C Talmadge A Tubis G Long and P Piskorski Model ing otoacoustic emission and hearing threshold fine structures Journal of the Acoustical Society of America Vol 104 No 3 September 1998 C Talmadge G Long A Tubis and S Dhar Experimental confirmation of two sources interference model for the fine struc ture distortion product otoacoustic emissions Journal of the Acoustical Society of America Vol 105 No 1 January 1999 M L Whitehead M J McCoy B L Lonsbury Martin and G K Martin Dependence of distortion product otoacoustic emissions on primary levels in normal and impaired ears I Ef fects of decreasing L2 below L1 Journal of the Acoustical Society of America Vol 97 No 4 April 1995 M Whitehead M McCoy B Lonsbury Martin and G Mar tin Dependence of distortion product otoacoustic emiss
15. Members of the group Felix Kochend rfer Marina Torrente Number of copies 6 The analysis of the results and their com parison with simplified cochlear models did Number of pages 114 not provide a universal tool for an estima tion of the threshold from the DPOAE fine Attachments structure and level 1 CD ROM Completed 02 06 2010 The contents of this report are freely available but publication with specification of source may only be done after arrangement with the authors Authors This report is written by Felix Kochendorfer felik es aau dk Marina Torrente marinatr es aau dk III Preface This report documents the Master Thesis written by group 1061 of the Sec tion of Acoustics Department of Electronic Systems at Aalborg University during 4th semester of the acoustics master programme of Aalborg Univer sity The purpose of this project is to examine the relationship between the fine structure of DPOAE and those of the hearing threshold The documentation presented in this report is structured as described next First an introductory chapter explaining the motivation and problem statement of the present research is presented in Chapter The necessary background theory related to the human auditory sys tem is presented in Chapter 2 to support further explanation of the hearing threshold and OAE fine structure nature which is given in Chapters 3 and Possible similar
16. The absolute threshold values vary from 0 to 30 dB HL among all the sub ject s results For most of the subjects the absolute threshold fluctuates around 10 or 15 dB HL APPENDIX A STANDARD AUDIOMETRY TESTS JOURNAL Subject 1 Right ear dB HL 2000 Frequency Hz Subject 3 Left ear 4000 dB HL x X X X x 2000 Frequency Hz Subject 5 Left ear 4000 dB HL 2000 Frequency Hz dB HL dB HL dB HL Subject 2 Left ear 2000 Frequency Hz Subject 4 Right ear 4000 X X X X x M UM 2000 Frequency Hz Subject 6 Left ear 4000 2000 Frequency Hz Figure A 2 Standard audiometry results from subjects 1 to 6 4000 A 4 Results dB HL dB HL dB HL Subject 7 Left ear 2000 4000 Frequency Hz Subject 9 Right ear 2000 4000 Frequency Hz Subject 11 Right ear 15K 20 257 30 X X X X X X KK XX 2000 4000 Frequency Hz dB HL dB HL dB HL Subject 8 Right ear 2000 4000 Frequency Hz Subject 10 Right ear X x X x x x 2000 4000 Frequency Hz Subject 12 Right ear X X X KK 2000 4000 Frequency Hz Figure A 3 Standard audiometry results from subjects 7 to 12 High Resolution Audiometry Tests Journal The high resolution audiometry test was performe
17. spectively The average frequency shifting of the frequencies corresponding to the maximum ripple level individually between the test retest thresholds is 1 68 octaves or 1 01 96 with a standard deviation of 1 67 octaves or 1 03 96 The ripple spacing deviation is 1 28 octaves for the average across subjects with a standard deviation of 1 30 octaves between subjects The ripple The average correlation coefficient is calculated by applying Fisher s transformation to the correlation coefficients calculating the arithmetic mean of the transformed coefficients and transforming it back by the inverse Fisher s transformation Heis 08 The equally judge frequency ranges are the frequency ranges which classification as fine structure or not is coincident in the test and retest measurement 50 11 3 Relationship between DPOAE and threshold fine structure Subjects 1 2 3 11 All Average octaves 1 81 1 179 1 113 1 38 1 68 Standard deviation octaves 1 339 1 162 1 2450 1 52 1 67 Table 11 3 Frequency shifting in octaves of the frequencies corresponding to the maxima of the threshold ripples when comparing the coincident ripples in the test retest results Subjects 1 2 3 11 All Average octaves 1 25 1 69 1 18 1 20 1 28 Standard deviation octaves 1 57 1 50 1 41 1 26 1 30 Table 11 4 Spacing deviation in octaves of the coincident ripples in the test retest results
18. 1 1 Comparison of the models with individual results The individual results from the experiment were analyzed with respect to the cochlear models In order to see similarities between the measurements and the models an excitation function E w has been developed for each subject based on the overall shape of the threshold determined with the high resolution audiometer The Parameters Rao Ry and 24 were adjusted individually to find the best correlation with the experimental results The main purpose of this analysis was to check if the appearance of ripples at specific frequencies is compatible with the models An example for a simulated DPOAE and threshold fine structure for Sub ject 1 is given in Figure Similar to the measured results the models show the highest DPOAE ripples in the range 1800 Hz lt f lt 2600 Hz Furthermore the DPOAE level increases while the fine structure decreases for 2f fo gt 2200 Hz DPOAE amplitude Threshold amplitude DPOAE amplitude Threshold amplitude 1600 2500 1600 2500 otc f 12 2 20 20 20 20 g 10 108 g 10 105 m om E S 2 5 o 0 0 0 0 E as E Z a a a E 2 E 5 10 10 8 E 10 10 8 E 22 2 2 32 ES 20 20 E 5 20 20 E 30 i 30 30 i i 30 1600 2500 1600 2500 2 4 f Figure 12 2 DPOAE black and threshold blue fine structure simula tions as a function of 2f f2 and f top left and right of Subject 1 For compariso
19. 4 fo left f middle and Vif right The errorbars represent the standard deviation between subjects 98 D 3 Ripple height Ripple spacing Hz Ripple spacing Hz Ripple spacing Hz 400 400 350 350 350 300 300 300 250 250 250 200 200 200 150 150 150 100 100 100 Soo 2000 4000 00 2000 4000 Soo 2000 4000 2f Hz f Hz 1 1 Hz Figure D 17 Average of the ripple spacing across the subjects in Hz in 1 8 octave bands for the threshold blue and DPOAE black fine structure as a function of 2 1 fa left fa middle and Y f f2 right The errorbars represent the standard deviation between subjects D 3 Ripple height The average and standard deviation of the threshold and DPOAE ripple height calculated for each subject within the frequency range from 1400 Hz to 2800 Hz is presented in Figure C Threshold EHH DPOAE as function of fio IE DPOAE as function of f MN DPOAE as function of f 1 Ripple height dB SPL 5 6 7 8 9 10 11 12 Average Subjects Figure D 18 Ripple height in dB SPL from each subject The threshold and DPOAE ripple height is calculated in 1 8 octave bands for each subject and averaged over all the subjects as shown in Figure 99 APPENDIX D THRESHOLD AND DPOAE FINE STRUCTURE COMPARISON OF RESULTS Ripple height Ripple height Ripple height 0 dB SPL E dB SPL a 8 dB
20. 4 in Figure where the minima 1 and 2 would fulfill the initial requirement regarding the maximum marked with an X but obviously those extreme values do not represent a ripple 39 CHAPTER 9 FINE STRUCTURE DETECTION ALGORITHM Once a maximum is neglected the whole procedure continues at the next maximum at a higher frequency If the maximum fulfills all requirements the remaining left and right minima are analyzed further in pairs of all possible combinations An ideal ripple is considered to have following characteristics e the frequency of the maximum is centered between the frequencies of the two adjacent minima e both adjacent minima have the same level e the height of the ripple is as high as possible Hence the pairs are analyzed regarding those three criteria The two mini ma which are closest to this ideal ripple are accepted and characterize a ripple together with the current maximum 9 2 Modifications for DPOAE fine structure de tection For the detection of fine structure in DPOAE measurements the algorithm is slightly modified compared to the threshold fine structure detector The ILO system analyzes additionally the noise around the distortion product It is represented by two curves The lower curve indicates one standard deviation and the higher two standard deviations from the background noise thus specifying the limit of the 95 96 confidence region According to the algorithm proposed by Reuter and
21. A the equipment test setup and procedure used to perform the standard audiometry are described The test lasted approx 10 minutes The second test was the high resolution audiometry Its aim was to screen the subject s threshold fine structure The procedure and setup of the test which lasted approx 20 to 25 minutes is described in detail in Appendix The third test was the DPOAE fine structure measurements This test did not imply the active participation of the subjects The subject s DPOAE fine structure was measured with the ILO96 system as described in Appendix The test had a duration of approx 20 30 minutes In order to test the reliability of the implemented high resolution audiome ter a subset of 4 subjects 2 males and 2 females performed the high res olution audiometry test twice The first test was conducted at least one week before the retest which was performed together with the standard audiometry and the DPOAE measurements The reasons for conducting the test retest threshold screening experiment in different days was entirely practical maintaining the duration of the main experiment to less than one and a half hours 10 3 Analysis strategy The analysis of the results is divided in three parts e Comparison of Standard and High resolution audiometry results e High resolution audiometer reliability e Relationship between DPOAE and threshold fine structure The strategy used in each part of the analysis is des
22. Clev 79 Cran 08 Elli 58 Fast 07 Fitz 04 Han 89 Harr 89 He 97 Heis 08 Bibliography W S Cleveland Robust Locally Weighted Regression and Smoothing Scatterplots Journal of the American Statistical Association Vol 74 No 369 December 1979 J L Cranford Basics of Audiology From Vibrations to Sounds Plural Publishing 2008 E Elliott A Ripple Effect in the Audiogram Nature April 1958 H Fastl and E Zwicker Psychoacoustics Facts and Models Springer 3rd Ed 2007 M J Fitzpatrick Real Time implementation of a hearing test ing technique Master of science Department of Electrical and Computer Engineering Clarkson University 2004 L A Han and T Poulsen Equivalent Threshold Sound Pres sure Level for Sennheiser HDA 200 Earphone and Etymotic Re search ER 2 Insert Earphone in the Frequency Range 125 Hz to 16 kHz Scandinavian Audiology Vol 105 No 12 1989 F Harris B Lonsbury Martin B Stagner A Coats and G Martin Acoustic distortion products in humans System atic changes in amplitude as a function of f2 fl ratio Journal of the Acoustical Society of America Vol 85 No 1 January 1989 N ji He and R A Schmiedt Fine structure of the 2 f1 f2 acoustic distortion product Effects of primary level and fre quency ratios Journal of the Acoustical Society of America Vol 101 No 6 June 1997 S Heise J Verhey and M Mauerman
23. SPL 8 2000 4000 boo 2000 4000 too 2000 4000 2f f Hz f Hz t Hz Figure D 19 Average of the ripple height across the subjects in dB SPL in 1 8 octave bands for the threshold blue and DPOAE black fine structure as a function of 2f fo left f middle and fif2 right The errorbars represent the standard deviation between subjects 100 High Resolution Audiometry Calibration and Noise Estimation E 1 Setup and Procedure The complete high resolution audiometer setup needed to be calibrated ac cording to ISO 389 After developing the basic algorithm and setting up all the required equipment the calibration was performed in the audiometric cabin B4 103 and the Control Room A B4 105 The equipment listed in Table E I was used in addition to the setup specified in Appendix B according to ISO 389 Model Type AAU Room no Artificial ear IEC 60318 with adapter DB 0348 0 5 measurement microphone mounted in B amp K 4153 Br el amp Kjar 4153 07631 B4 103 Br el amp Kj r 4134 08129 B4 103 Br el amp Kj r 2669 Microphone preamplifier 56509 B4 103 Br el amp Kj r 2636 Measuring amplifier 08451 B4 105 Br el amp Kjaer 2807 Microphone power supply 07305 B4 105 Br el amp Kj r 4230 Sound level calibrator 08373 B4 103 Toshiba T3200SX Notebook with installed 26855 B4 105 with MLSSA Acoustical me
24. a specific frequency range Ty is the less consistent within the thresholds from the two first repetitions The cross fade is applied to the four previous and subsequent frequency points regarding the remeasured frequency range limit Since the third repetition is not performed outside the specific range the limit value is used in the calculations dB SPL 0 d 1500 1550 1600 1650 1700 1750 1800 1850 1900 Frequency Hz Figure 7 2 Example of a cross fade between thresholds in the limit of the frequency range of the third repetition The final threshold is calculated as the average between the two resulting curved Figure shows an example of the whole procedure In upper figure the standard deviation of the normalized difference between the two thresholds X is compared with the limit value of 2 5 dB In the frequency ranges where gt exceeds the limit the threshold is remeasured In the middle figure the three measurements are represented Finally the figure in the bottom shows the two most consistent thresholds with a smoothed transition in the overlapping regions and the final threshold as the average of the former The Matlab function two_closest_thresholds m included in the enclosed CD provides the two more consistent cross faded thresholds 32 7 2 Repetitions and consistency checks Threshold dB SPL DER 1 repetition 24 repetition J 3 repeti
25. agrees with the analyzed models where the rapidly varying phase of the apical reflectance is assumed to be the reason for the periodicity This phase influences the spacing between fine structure minima and maxima in both models in the same way Despite the observed similarities and agreements with the models a univer sal tool for an estimation of the threshold from the DPOAE fine structure and level could not be provided 69 Future work With the purpose of studying a possible relationship between their fine struc tures the threshold and DPOAE were screened Measurements of the hear ing threshold require of concentration of the subjects and therefore the time duration of this test should be kept short This limits the frequency range of the measurement However measurements in a wider frequency range will allow a more complete comparison of fine structures since both threshold and DPOAE are very frequency dependent Hence the design of a test pro cedure that allows a longer measurements time without biasing the subjects performance could be studied Although the implemented high resolution audiometer is a fast and precise method for tracking threshold fine structure it is not accurate on the es timation of absolute threshold level Therefore it could be interesting to analyze and implement other high resolution audiometric methods and test their performance for screening hearing thresholds The screening of DPOAE is highl
26. cells are the most sensitive to damage caused by high noise levels The active process of the outer hair cells also causes emissions of sound from the cochlea known as otoacoustic emissions Otoacoustic emissions are often observed as a result of a distortion product when the cochlea is excited with two sinusoidal sounds When recording those distortion product otoacoustic emissions DPOAE quasi periodic level variations with minima and maxima can be observed known as DPOAE fine structure The presence or absence of this fine structure is considered to be an indication of the status of the outer hair cells A similar fine structure can also be found when determining the hearing threshold with a very high frequency resolution This might be a result of similar mechanisms which are responsible for the fine structure of the DPOAE Hence this small variations in the hearing threshold curve which can be up to 15 dB are not an indication of hearing disorder The purpose of this project is to examine the relationship between the fine structures of DPOAE and those of the hearing threshold For this purpose the hearing threshold and the DPOAE of the same subjects have to be determined with a high frequency resolution With a classical audiometry this fine structure cannot be observed due to the low frequency resolution The audiometric test is always dependent on the subject s response and requires a high concentration An increase of 5
27. correlation representing the human average the subjects were not screened beforehand for any special threshold patterns In general the determined fine structures show a good agreement with re ported fine structure data by other authors However due to the detection algorithm the results are biased in a certain way Since there is no clear definition whether variations in DPOAE or threshold are considered as fine structure or not the limits of the detector were adjusted according to pre vious investigations Hence all parameters like ripple spacing height and prevalence are influenced by initial restrictions for the classification of a ripple The obtained data of DPOAE and threshold fine structure shows on average results which coincide with the theory and the models The ripple spacing in octaves just shows very little variations among frequency and both threshold and DPOAE fine structure indicate a common periodicity Similarities with previous investigation of different authors can also be found regarding the ripple height In general the fine structure of DPOAE is more pronounced and shows a higher level difference between adjacent maxima and minima According to the theory more parameters are involved in the generation of DPOAE than in the threshold fine structure hence the higher variation in the ripple height of the DPOAE is comprehensible Additionally a higher prevalence of the fine structure is observed on average for DPO
28. criterion of ALmin are used for further analysis In the following minima at frequencies lower than fmar will be referred to as left minima according to their position in the graph regarding the maximum and min ima at frequencies higher than fmax as right minima For the selection of those minima which can be considered to represent a ripple the left and the right minima are analyzed individually Figure 9 1 presents two examples for the detection of minima and maxima as fine structure elements The X denotes the maximum under current analysis Minima in the range fmaz Afmaz fmaz Afmax are marked with a circle Whether an extreme value is accepted or not is explained in the following Minima that are more separated in frequency from finas than closer minima presenting a lower level are neglected since they do not represent an absolute minimum in the range of a ripple Furthermore if two minima are separated by a maximum which fulfills the ALmin criterion the mini mum with a higher distance in frequency regarding fmaz is neglected in the current analysis as well since it is considered to belong to a new ripple 38 9 1 Detection procedure This is illustrated in the threshold curve of Subject 12 in Figure 1 Minima 1 and 2 are considered to be left and 3 and 4 right minima Minimum 1 is not considered as a possible minimum representing a ripple around fj because minimum 2 appears at lower level whereas m
29. f and f overlap The energy of 2f fo is transmitted to its characteristic tonotopic place in the BM where is reflected backwards to the base As stated in 3 3 theoretical models assume that the roughness of the BM is responsible for the apical reflectance that produces the threshold and DPOAE fine structure According to Talmadge et al Talm 99 for a constant f f ratio the distortion compo nent will present a short latency and small phase variation with frequency whereas the reflection component will present long latency and its phase will vary rapidly as a function of frequency The generation of the DPOAE fine structure will be due to this phase difference between the two components The variations in the phase of the reflection component are assumed to de pend on the phase of the apical reflectance which varies with wap 27 fap or the tonotopic place equivalently Hence if the phase of the distortion com ponent is assumed to be constant the ideally periodic DPOAE fine structure will be generated as illustrated in Figure due to the interference of the two components It has been suggested among the literature that the presence of fine structure is an indicator of a healthy ear Reut 06 The experiments and simulations described in Maue 99 were conducted to analyze the DPOAE fine structure in subjects with different kinds of known hearing impairment It was ob served that when the reflection component fap was in the fr
30. has been reported in the literature suggesting a common origin in their mechanism of generation Hence an experiment was conducted where the threshold and DPOAE fine structure of different subjects were measured and the results were analyzed in order to find a relationship between them The DPOAE measurements were carried out using a calibrated commercial system see Chapter 7 and the threshold fine structure was tracked with the implemented high resolution audiometer see Chapter 5 To be able to rely on the determined threshold curves correspondent validation tests were necessary A comparison with a standard audiometry results and a test retest were conducted for this purpose 41 CHAPTER 10 EXPERIMENT DESIGN Direct comparison of the threshold and DPOAE curves is complex and not always possible A thorough analysis of individual and averaged results should be regarded instead Thus the fine structure detector implemented see Chapter 9 was applied to objectively study the relationship between the presence periodicity and level of the threshold and DPOAE ripples Theoretical models of the cochlea have tried to explain the mechanisms of generation of fine structure in the threshold and DPOAE Comparison of these models with the experimental results can be useful to find out a possible relationship between both generation mechanisms 10 2 Procedure The experiment was carried out in the Acoustics Laboratory in Aalborg
31. hearing threshold in the cases of subject presenting cochlear hearing loss Part I Appendices Standard Audiometry Tests Journal Prior to the DPOAE measurement and the high resolution audiometry a standard pure tone audiometry test which complies with ISO 8253 1 was performed Since threshold tracking methods are not expected to be precise absolute threshold estimates the aim of this test was to obtain an estimate of the absolute hearing threshold level at a discrete number of frequencies Fur thermore the data obtained from the standard audiometry test was used to check for possible hearing damage in the analyzed frequency range The audiometric test was performed using the ascending method introduced by Carhart and Jeger 1959 Lydo 99 in the range from 1250 to 4500 Hz with a resolution of 6 points per octave A 1 Setup The test was performed in the Audiometric Cabin A room B5 102 in the Acoustics Laboratory at Aalborg University Fredrik Bajers Vej 7 9220 Aal borg E Figure A 1 shows the setup of the clinical audiometer During the test the equipment and the operator were in the same room as the subject It is important to ensure that the subject is not distracted by the operator or any indications of the equipment during the test This was simply done by turning the chair where the subject was seated towards the wall A 2 Equipment The equipment used for the standard audiometry test is detailed in Table
32. in Table Model Type AAU Room no Siemens PC 43919 B4 103 Otodvitamitos Otoacoustic emissions measuring ILO9G system including software am 52661 B4 103 plifier unit and ear probe Table C 1 DPOAE measurements setup equipment The ILO OAE software is a DOS application which runs on MS Windows 98 The software analyzes the data extracting the DPOAE amplitude spectrum and the noise by the use of internal algorithms The card interface of 2x2 channels generates the stimuli signals and receives the recorded signals from the microphone It is also responsible of controlling the internal switches and attenuations The amplifier unit amplifies and or attenuates the output input signals generated stimuli and received signals The probe is connected to the amplifier unit by an eight pin DIN connector C 3 Test Procedure To screen the DPOAE fine structure within the frequency range specified in Section B 1 3 several measurements were required The DP gram test of the ILO96 system was set to a micro resolution This provides measurements within a range of 200 Hz for fo 3 kHz and of 400 Hz for fo gt 3 kHz presenting 17 pairs of primary tones Hence 13 concatenated measurements were required to cover the desired frequency range Prior to each measurements the program executes a probe checkfit It uses a click stimulus to measure the frequency response of the ear canal detecting anomalies in the
33. in the artificial ear can be calculated with the sensitivity of the B amp K 4134 A detailed sketch of the calibration setup is given in Figure E 1 Control Room A Audiometry Room B4 105 B4 103 Amplifier Pioneer A 616 Headphone Sennheiser HDA 200 E E ox Artificial Ear B amp K 4153 Measuring Amplifier B amp K 2636 Preamplifier B amp K 2669 Figure E 1 The equipment its connections and position used for the cal ibration of the high resolution audiometer 102 E 2 Calibration Data E 1 3 Noise Estimation The noise floor of the equipment was estimated by measuring the voltage at the terminals of the headphone while no signal was played with the setup It was analyzed in third octave bands with the B amp K 2231 In order to obtain a reference to compute the sound pressure level a 1 kHz signal with a gain g 0 1 was played while measuring the voltage at the terminal with the same setup in a third octave band centered at 1 kHz The measured voltage of this signal was 2 21 mV which can be referred to a sound pressure level of approx 63 dB SPL With the data from the calibration and the headphone frequency response the apparent sound pressure level of the noise at the ear could be estimated A detailed sketch of the noise estimation setup is given in Figure Control Room A Audiometry Room B4 105 Wall B4 103 Headphone HDA 200 40dB
34. kHz to 1 9 kHz The resulting average correlation coefficient is 0 84 This indicates that in general the audiometer provides a good reproducibility of the shape of the threshold whereas the level is not necessarily precise Since the audiometer is implemented to screen fine structure in the threshold it is important to check how precisely the distribution of peaks and valleys is reproduced Therefore the ripple parameters obtained by ap plying the fine structure detector to the two determined curves are analyzed The results of the analyzed prevalence of threshold fine structure in the test retest measurements are presented in Table 11 2 The average difference between the percentage of ripple prevalence in test and retest is 13 9 whereas the average of equally judge frequency ranged is 74 2 96 showing a good reproducibility of the audiometric system regarding ripple prevalence Subjects 1 2 3 11 1 Test 65 7 79 8 19 2 77 8 2nd Test 38 4 73 7 39 4 75 8 Equally judge ranges 72 6 69 7 79 8 77 8 Table 11 2 Percentage of threshold ripple prevalence in each measurement first and second rows and percentage of equally judge fre quency ranges comparing the two measurements third row The shifting of frequency regarding the absolute maximum of the ripples and the spacing and height deviation of the coincident detected ripples in the test retest thresholds are presented in Tables 11 3 and re
35. probe fit The data from the checkfit is used to balance and normalize the two primary stimulus levels The spectrum analysis is performed by the system applying Fast Fourier Transform FFT with a 12 2 Hz resolution The noise is estimated taking the 10 closest to 2 1 f components in the FFT excluding the distortion 85 APPENDIX C DPOAE FINE STRUCTURE MEASUREMENTS TESTS JOURNAL product itself The system presents two noise curves The lower curve indicates one standard deviation and the higher two standard deviations from the background noise thus specifying the limit of the 95 confidence interval The measurements procedure followed to carry out the DPOAE measure ments is described next 1 Start the program ILO OAE 2 Select the option D DPOAE 3 Click on Menu Setup Load parameters and load the file DP_ PARAM PAR in the current directory This will set the primary level to L L2 65 45 dB SPL 4 The next steps are repeated for each of the 13 measurements of 0 1 octave to cover the desired frequency range e Select the DP gram test in Menu Test select DP gram or press F6 e Chose the Micro structure 0 1 oct resolution e Enter the correspondent central value of fo in the range of the current measurement e Perform the probe checkfit If the fitting is not successful the spectrum will present frequency regions colored in red and the test stimulation will not be o
36. structure might vary depending on the measured frequency range Hence and also due to practical limitation the frequency range for measuring DPOAE fine structure needs to be defined 8 1 1 Primaries frequency ratio Harris et al Harr 89 studied the influence of primaries frequency ratio f fi on the DPOAE amplitude It was concluded that a ratio of fo fi 1 22 provides the largest DPOAE amplitude regardless frequency and level Therefore this is the value of the primaries ratio chosen for the DPOAE fine structure measurements 34 8 2 ILO96 8 1 2 Primaries level choice Decreasing La below Ly when Li lt 75 dB increases the DPOAE amplitude This effect depends on the value of L and on the frequency Whit 95a The choice of primary levels L L2 65 45 dB is made according to the discussion presented by Reuter and Hammershgi Reut 06 It was based on the experimental results presented by Whitehead Whit 95b where different level combinations L L2 where tested at three geometric mean frequencies 1 39 2 79 and 5 57 kHz with f f 1 21 This study revealed that the combination of primary levels L1 L3 65 45 dB provided relatively high DPOAE amplitude and it appeared to be a good compromise for measuring presence of fine structure 8 1 3 Frequency range As stated by Reuter and Hammersh i Reut 07 the prevalence of the fine structure is stronger in the mid frequency range from 1 kHz to 3 kHz It was claim
37. t l O m N em t l e T EN 52 t l e T EN m t e e e e o o o o LO ive LO LO e e e N a E m ro TN re TL a M el a o o o H o o amp amp z amp amp on E om E o o o a o gt o o o o o o o o o o o o o o o o o o m v qd T T To P O T mS le cms ub TO Y o7 Y 1ds gp epnijdure 3voda Das gp epnyjdwe 3voda Ids ap epnyjdwe 3voaa Kas gp spmdwe 3voda Threshold and DPOAE fine structure of subject 12 1 2 f 5 Figure D 12 96 D 1 Ripple prevalence D 1 Ripple prevalence The average and standard deviation of the threshold and DPOAE ripple prevalence calculated for each subject within the frequency range from 1400 Hz to 2800 Hz is presented in Figure 100 90 Threshold E DPOAE as function of fip IE DPOAE as function of f MI DPOAE as function of 1 80 70 60 50 40 30 20 Ripple prevalence 96 5 6 7 8 9 10 11 12 Average Subjects Figure D 13 Average ripple prevalence for each subject The threshold and DPOAE ripple prevalence is calculated in 1 3 octave bands for each subject and averaged over all the subjects as illustrated in Figure Ripple prevalence Ripple prevalence Ripple prevalence 1 100 100 80 80 80 60 _ 60 _ 60 x z zx 40 40 40 20 20 20 0 0 0 1000 2000 4000 1000 2000 4000 1000 2000 4000 2t A Hz t Hz 5 Hz Figure D 14 Average of the ripple prevalen
38. test retest high resolution audiometry was conducted with four of the subjects see Chapter 10 in order to check the reliability of the implemented high resolution audiometer The results of the test retest measurements are summarized in Figure The correlation coefficient r see Equation 10 1 gives information regarding the correlation between the output thresholds from the two experiments in terms of overall shape Reliability test subject 1 46 r 0 93 dB HL i i i I dd 1600 2000 2500 Frequency Hz Reliability test subject 3 15 46 r 0 89 dB HL L 1600 2000 Frequency Hz dB HL dB HL 1 run threshold 2 run threshold 1 run ripples 2 run ripples Reliability test subject 2 l 1600 2000 2500 Frequency Hz Reliability test subject 11 1600 2000 Frequency Hz Figure 11 3 Test retest of the high frequency resolution audiometer based on the FINESS algorithm results 49 CHAPTER 11 RESULTS Subject 1 presents the best correlation between the shape of the two thresholds despite the level differences On the other hand Subject 11 shows the weakest correlation but there are no significant differences re garding the absolute level The lack of correlation is this case appears due to differences in the shape of the thresholds mainly in the frequency range from 1 5
39. the subjects in Group 3 Since the threshold shows a higher prevalence of fine structure than the average high values for either Ra or R are expected On the other hand the DPOAE measurements show just very little fine structure 63 CHAPTER 12 COCHLEAR MODEL SIMULATIONS but a generally high level This may be modeled by assuming a strong imbalance between aqist and aref1 leading to an either very high or low value for R The overall shapes of the threshold and the DPOAE are more similar when the DPOAE is plotted with respect to 2 f4 fo as shown in F igure 12 4 This assumes a direct influence of R on the DPOAE level Hence a high value for Rao returns the best results Threshold amplitude DPOAE amplitude Threshold amplitude DPOAE amplitude 1600 2500 1600 2500 20 20 20 20 DPOAE amplitude dB SPL I 3 3 DPOAE amplitude dB SPL 1 3 S Threshold amplitude dB SPL Threshold amplitude dB SPL l mm o 20 eo o 30 1600 2500 1600 2500 att f Figure 12 4 Comparison of modeled top and measured bottom DPOAE and threshold fine structure of Subject 7 Group 3 Blue lines indicate the threshold and black lines DPOAE The DPOAE fine structure is reduced when the excitation in creases at 2f f2 which is a result of applying a relatively high value for Ral Group 2 represents the majority of the all subject
40. three cavities scala vestibuli scala tympani and scala media The cavities are separated by two membranes the Reissner s and basilar membranes BM and they extend all along the cochlea The part of the cochlea closer to the middle ear is known as the base and the tip of the cochlea is known as the apex NS iV qu tectorial y 4 membrane I spiral limbus cochlear nerve spiral fibers ligament aa i TUE A PS basilar Ra ed tk XM 5 membrane eres tympani organ of Corti Figure 2 2 Cross section of the cochlea divided into the three fluid filled cavities scala vestibuli scala media and scala tympani Re drawn from source http media 2 web britannica com eb media 01 14301 004 4B6F34DA gif The scala vestibuli terminates in the oval window and the scala tympani in the round window They are filled with a fluid called perilymph The perilymph can flow thanks to an opening in the apex of the cochlea the helicotrema Piston like movements of the oval window set the perilymph in motion producing traveling waves along the Reissner s and basilar mem branes The hair cells present in the organ of Corti see Figure 2 3 are responsible for the generation of neural impulse responses that are conducted via the auditory nerves to the central auditory nervous system a CHAPTER 2 THE HUMAN AUDITORY SYSTEM Scala media Tectorial membrane Outer hair cells A V Efferent fibre Afferent P
41. to be a result of reflections of the traveling waves inside the cochlea Following this hypothesis the incoming wave traveling from the base towards the apex is reflected in a certain frequency dependent region along the basilar membrane Kemp 79 This is commonly referred to as the apical reflection Zweig and Shera Zwei 95 considered that this reflection is due to random inhomogeneities or roughness near the tonotopic location see Section 2 3 1 This roughness is assumed to be the reason why the phase of the reflectance is strongly dependent on frequency Due to an imperfect impedance matching at the stapes the retrograde tra veling wave will be reflected again resulting in an overlay of the reflected and incoming wave Their phase difference which is mainly caused by the apical reflectance at the tonotopic location will cause a constructive and destructive interference of the waves and thus either enhance or reduce the amplitude of the traveling wave Hence the rapidly changing phase of the apical reflectance causes an uneven distribution of the energy with minima and maxima for different frequencies Talm 98 Heis 08 Comparisons between analogue model computations and fine structure mea surements predict in some cases that the apical reflectance is very high and even exceeds 1 Kemp 79 The active process of the outer hair cells see Section 2 3 2 is therefore assumed to be involved in the reflection With increasing stim
42. w Hearing Level dB HL 10 Pi LANE 40 X 1 1 r 1 po poc 125 250 500 1k 2k 3k 4k 6k 8k 125 250 500 lk 2k 3k 4k 6k 8k Frequency Hz Frequency Hz Figure 3 2 Changes of the absolute hearing threshold up to 40 dB pre sented as physical sound pressure level left and in a clinical audiogram right The dashed line represents the reference according to ISO 389 Especially young people might present a negative absolute threshold in dB HL This simply indicates that their hearing is more sensitive than the average A 3 3 Audiometric fine structure When determining the absolute hearing threshold with a very high frequency resolution quasi periodic level variations can be observed in the audiogram The differences between adjacent peaks and dips can thereby be up to 15 Pu 3 3 Audiometric fine structure dB Heis 08 These ripple effects have been discovered first by Elliott in 1958 who already assumed this as a universal phenomenon He demonstrated that fine structure as it is specified nowadays is stable over a long period of time but differs from ear to ear Apparently fine structure can only be found in frequency ranges where the subject does not show signs of a hearing impairment On the other hand listeners without a detected fine structure do not necessarily show a hearing loss Although the origin of the threshold fine structure is not completely under stood it is assumed
43. when 2 f f falls into an impaired region whereas the average DPOAE level remains constant The DPOAE level fluctuates 60 12 1 Analytical models around an average level of aqist Since aqist is influenced by the BM condition at the tonotopic location of f2 the overall level changes when f falls into the impaired region Experiments by e g Mauermann Maue 99 show similar results The threshold fine structure disappears almost completely in this model since R is reduced in the impaired region Hence the ripple height is directly related to Rq in both DPOAE and threshold fine structure as long as Ra lt 1 A different behaviour occurs when R exceeds 1 In this case the overall DPOAE level is also influenced by the reflection component as illustrated in Figure 12 1p The maximum apical reflectance is set in this example to Rao 3 This results in a reduction of the DPOAE fine structure due to the unequal levels of ar and aqist since the ripples are expected to be highest when Gre fi agis Once 27 f falls into an impaired region resulting in a decrease of Ra the DPOAE level is reduced at first while the ripple height increases When R falls below 1 the DPOAE show the same behaviour as presented in Figure 12 1p Due to the initially high reflectance Rao the fine structure of the hearing threshold is more pronounced in the second example CHAPTER 12 COCHLEAR MODEL SIMULATIONS 12
44. 2 A Standard Audiometry Tests Journal 73 Gee hes bf oh ores a Gees ae ares Sets ee CM 73 Fos ade Ros Be ae Seo Ge ge Se ee ee eee Ges 73 A 3 Test Procedure ees 74 AA Results m s es oak 4s Xo he komo a Rx E ded 75 B High Resolution Audiometry Tests Journal 78 A aa a a ee eee da 78 O A Pm 79 B 3 Test Procedure o 80 BA Results scene aos sky e EC ode ee vos 81 CONTENTS 84 A a sk aces mo oe ee At xD poh de RE Rec ins ean 84 C 2 Equipment 0002 ee ee 85 mum er Bonn oes Soak EUER E MUS RE RENE 85 CA Results eie eo eek ia A ee E we eG ag 87 D Threshold and DPOAE fine structure comparison of results 90 eee ee ee TE ae a PE 97 be ack Hanh ation a Gee Gnu ogee geek ce 98 ee ee aaa are 99 101 od ee eee Ge ee ee 101 SA ey SES PS 102 Be are Sia ey ee 102 res M Be ee Wk a ER 103 bes eae oe eee Gee ee ee 103 eee ee ee dee e 104 oe c 104 ea Ghee A Ee Ge eee ge 104 hese dee ad 106 107 109 Bibliography 111 El Introduction and problem statement In the human auditory system the cochlea plays a major role in the cognition of sound The outer hair cells act as a mechanical amplifier increasing the motion of the basilar membrane This stimulates the inner hair cells which transform the vibrations into neural impulses Due to the active process of the outer hair cells even very low signals will be amplified which highly influences the absolute hearing threshold However the outer hair
45. 3 2 The audiogram A Introduction and problem statement 2 1 Outer ear 2 2 Middle ear 2 9 Inner ear 2 3 1 Traveling waves 2 3 2 The outer and inner hair cells function 3 3 Audiometric fine structure Otoacoustic emissions 4 1 Physiology behind the generation of OAE 4 2 OAE measurement techniques 4 3 DPOAE fine structure Contents 5 Models of DPOAE and hearing threshold fine structure 21 CONTENTS 24 6 1 Level presentation strategy lees 24 6 1 1 The B k sy method o 25 6 2 Frequency sweeping strategy llle 25 6 2 1 The Audioscan method 25 6 3 Level tracking vs frequency sweeping 26 28 DRA A as 28 Side dte ee 30 8 DPOAE measurements 34 C in ee a se 34 job E nom A o E P be ee eS 34 vb bE DAS Ales E Rom po OR E 35 CCP p 35 B 2 MOOG cook e ee aE ee Ee ee a 35 37 oh ie ee ER ee ee el 38 Bet eye 40 41 Judge bp UR EPOR a HO eae RE 41 CIT 42 10 3 Analysisstrategy 2n 43 al roda 13 10 3 2 High resolution audiometer reliability 44 PER aa es ERE EIER EE 45 2 CONTENTS 46 TT Comparison of standard and o oda 46 jad ia 24 49 51 L0 aod dan es Pa a E E 52 11 3 2 Spacing rc a AA 53 1L3 8 Height s sa sece saaa sata aaa ee a 55 fe dote e g 57 59 a E EE RUN deep E 59 12 1 1 Comparison of the models with individual results 62 65 14 Conclusion 68 15 Future work 70 I Appendices 7
46. 73 APPENDIX A STANDARD AUDIOMETRY TESTS JOURNAL Clinical Audiometer Madsen Orbiter 922 2 Headphone Sennheiser HDA 200 Pushbutton Figure A 1 Setup used for the standard audiometry Model Type AAU Room no Clinical audiometer with push button Sennheiser HDA200 Audiometry headphone 52735 B5 102 Madsen Orbiter 922 33968 B5 102 Table A 1 Standard audiometer setup equipment The audiometer has a built in headphone amplifier hence the headphones can be connected directly to the headphone output It is important that the two channels are connected correctly Additionally the audiometer provides two external pushbuttons for each ear to record the subjects responses A 3 Test Procedure The algorithm of the test follows a method introduced by Carhart and Jeger It is a particular implementation of the ascending method and has been recommended by the American Speech and Hearing Association ASHA in 1978 Lydo 99 The algorithm starts at 40 dB hearing level and decreases with 10 dB steps 5 dB steps are used for the ascents to detect the hearing threshold A threshold level is validated after being detected two times at the same point The detection of the threshold runs automatically for each frequency but when analyzing in the resolution of 6 points per octave the start level and the frequency of each point to has to be set manually The procedure to run the standard audiom
47. AE compared to the threshold curves However regions with a high threshold prevalence do not necessarily show a high prevalence of DPOAE fine structure A surprisingly high similarity between the two prevalences analyzed in 1 3 octave bands can be found when presenting the DPOAE as a function of fo This result can not be explained with the models or previous studies since the reflection component generated at the tonotopic site of 2 4 f2 is assumed to be responsible for the appearance of fine structure whereas the distortion component rather seems to influence the overall DPOAE level 66 In contrast the differences in prevalences of threshold and DPOAE as a function of 2 1 f can be supported by the results of the model simu lations In cases for which the reflection component is stronger than the distortion component a reduction of the DPOAE fine structure is possible whereas the threshold fine structure remains or even increases Subjects 1 4 and 7 show a similar behavior Hence the comparison of those subjects results with the models may lead to the assumption that apical reflectances higher than 1 are possible This suggests that the cochlear amplifier is strongly involved in the generation of the DPOAE fine structure and may even compensate for differences due to the longer propagation distance of the reflection component The reflectances R applied in the models to achieve a maximum agree ment with the measure
48. Attenuator Artificial Ear 4153 Frequency Analyzer Figure E 2 The equipment its connections and position used for the noise estimation E 2 Calibration Data Date of Calibration 31 March 2010 Location Aalborg University Fredrik Bejers Vej 7 e Control Room A B4 105 e Audiometry Room B4 103 Sensitivity of Microphone in Setup 12 5 mV 103 APPENDIX E HIGH RESOLUTION AUDIOMETRY CALIBRATION AND NOISE ESTIMATION E 2 1 Gain Factors and Associated SPL Gain Factor g Measured Voltage associated SPL 0 4 1 3 mV 74 3 dB SPL 0 2 0 66 mV 68 4 dB SPL 0 1 0 33 mV 62 4 dB SPL 0 05 0 17 mV 56 6 dB SPL 0 025 0 09 mV 51 1 dB SPL The noise measured with the given setup was fluctuating at approximately 46 dB SPL Since these fluctuations were already visible for gain factors of 0 05 and 0 025 those measured values were neglected for further calculations E 2 2 Headphone Transfer Function The normalized transfer function TTE of the Sennheiser HDA 200 head phone used for the audiometric tests is given in F igure E 3 Headphone transfer function 20 T FFT dB o T 1 20 1 1 1 10 10 10 Frequency Hz Figure E 3 The headphone transfer function of Sennheiser HDA 200 nor malized at 1 kHz E 2 8 Noise Measurements Figure shows the electric noise measured in third octave bands at the terminals of the headphone
49. ESULTS DPOAE level and a better signal to noise ratio The ripple prevalence in the threshold curves is around 55 in both subjects see Figure 11 4 These three groups present very different characteristics Groups 1 and 3 are constituted only by two people each however they are of significant interest They represent examples of opposite phenomena presence of fine structure versus high DPOAE level Theoretical models of the basilar membrane have been implemented in order to analyze the different results provided by each group A description of this model and the results obtained are given in the next chapter Cochlear model simulations The results of the high resolution audiometry and DPOAE measurements are to be analyzed with respect to the theory presented in Chapters 3 3 For this purpose simplified cochlear models were implemented and applied to each subject s result with respect to the groups gathered in Chapter 11 12 1 Analytical models In order to analyze the individual data simulations of the cochlear behaviour have been implemented following simplified versions of the models by Tal madge et al described in Chapter 5 Considering an ideal uniform behaviour of the cochlea absolute values of all reflectances are independent of frequency two major cases have been studied to evaluate the DPOAE fine structure e the amplitude of the reflection component is higher than the amplitude of the distortion compone
50. Hammershoi Reut 06 ripples shall be rejected whenever a maximum is less than 3 dB above the limit of 95 96 confidence region However minima below this curve are still regarded for the analysis Furthermore ripples in the DPOAE fine structure often appear higher than threshold ripples and show typically very narrow notches characterizing a minimum Absolute maxima of ripples are not necessarily centered between the two minima hence the detection of the optimal pairs of minima is weighted more to ripple height than to an equally frequency spacing Experiment design In this chapter the motivation for conducting an experiment to screen thresh old and DPOAE fine structure is presented The experiment procedure is also described Finally the strategy followed for the analysis of the results is outlined 10 1 Motivation Otoacoustic emissions measurements are nowadays mainly employed for screening hearing loss providing an objective tool for hearing assessment However knowledge of the relationship between OAE and threshold curves is required in order to use OAE in clinical diagnosis As already stated fine structure can be observed in both threshold and DPOAE measurements Investigating the nature of the relationship between threshold and DPOAE fine structure is the focus of this project Both mea surements provide information about the active cochlear processes More over similar periodicity in the OAE and threshold fine structure
51. Scala tympani Figure 2 3 Organ of Corti in a cross section of the cochlea Redrawn from source http download videohelp com vitualis med organ of corti gif 2 3 1 Traveling waves Traveling waves produce a maximum vertical displacement at an specific place of the BM known as the tonotopic location showing a resonant be havior After reaching its maximum the wave amplitude decreases rapidly and dies near the apex Moor 07 The BM acts as a frequency analyzer At the base of the cochlea it is thin and narrow and it becomes thicker and wider as it approaches to the apex Hence high frequency tones will excite points of the BM close to the base whereas low frequency tones will excite points closer to the apex The tonotopic location is logarithmically dependent on frequency Fast 077 Figure shows the excitation pattern of the BM when it excited with a three tonal frequency signal Each tone produces a maximum of excitation in the BM corresponding to their characteristic tonotopic place This linear and passive mechanism which relies on the mechanical properties of the BM is not the only process involved in the generation of traveling waves A non linear active mechanism dependent on the operation of the outer hair cells also influences the BM traveling waves pattern in response to sound 2 3 2 The outer and inner hair cells function The outer hair cells OHCs contain muscle components This enables them t
52. Simplified models were implemented and compared with the results of the measurements in order to evaluate possible trends regarding cochlear pa rameters such as basal and apical reflectances An certain agreement of the results with previous studies and the models was observed However since the generation of the DPOAE fine structure involves more parameters than the fine structure of the threshold according to the models it is hard to establish a direct relation Regarding the average among all subjects the observed fine structure in DPOAE was in general more pronounced and showed a higher prevalence On the other hand presence of fine structure in the threshold did not neces sarily coincide with an increased ripple prevalence or height in the DPOAE and vice versa It appears according to the models that this is due to a high variation of the cochlear parameters involved in the fine structure generation across subjects It is hard to predict those parameters just from either the threshold or the DPOAE curve Hence a direct relation regarding the ripple height and prevalence between both DPOAE and threshold fine structure could not be determined High similarities were found in the periodicity instead The ripple spacing of the DPOAE and threshold fine structure can hence be considered to be of the same origin An average ripple spacing of approx 1 10 octaves was observed which showed just very little variations in frequency This strongly
53. University Fredrik Bajers Vej 7 9220 Aalborg E A total number of 12 subjects 7 males and 5 females between 20 and 31 years of age participated in the experiment Two of the subjects suffered of undiagnosed occasional tinnitus One subject had surgery at the age of 3 and another subject suffered from otitis during childhood All the subjects showed neither a severe hearing impairment nor an exceptional hearing sensitivity The experiment consisted of three tests 1 Standard audiometry 2 High resolution audiometry 3 DPOAE fine structure measurements Before starting the experiment the subjects were asked not to come directly from a noisy environment Thus at least during half an hour prior to the experiment they should stay in a relaxed and advisable quiet atmosphere The subject were suitably instructed before starting the tests see Appendix F They were also asked to fill in a questionnaire see Appendix G to ob tain general information about the subjects and their hearing status Breaks of 5 minutes were included between the tests in order to ensure the comfort of the subjects This way they could stay concentrated during the perfor mance of the tests when required The first test carried out in the experiment was the standard audiometry The standard audiometry was conducted to obtain an estimate of the ab solute threshold level at a discrete number of frequencies In Appendix _42 10 3 Analysis strategy
54. a AALBORG UNIVERSITY Marina Torrente Rodriguez Felix Christian Kochend rfer Helation between fine structures in hearing thresholds and distortion product otoacoustic emissions Master Thesis Section of Acoustics Spring 2010 Faculties of Engineering Science and Medicine Group 1061 Department of Electronic Systems a AALBORG UNIVERSITY Aalborg University Department of Electronic Systems Frederik Bajers Vej 7 9220 Aalborg Telephone 96 35 86 00 http es aau dk Title Relation between fine struc tures in hearing thresholds and distortion product otoacoustic Synopsis emissions The purpose of this project is to examine the Project term relationship between the fine structures of ACO10 spring 2010 DPOAE and those of the hearing threshold An experiment was conducted with 12 sub jects Their hearing threshold was screened Project group using an implemented audiometer with high Group 1061 frequency resolution and based on level presentation strategy Measurements of the distortion product otoacoustic emis sions DPOAE were also performed with a high frequency resolution A developed fine structure detector was applied to ex tract objective parameters characterizing the threshold and DPOAE fine structure A direct relation between the threshold and Supervisor DPOAE fine structure and level could not Dorte Hammersh i be found However similarities in the fine structure periodicity were observed
55. ar short duration tonal sounds of a certain level You must press the button as soon as you hear the sound and keep it pressed for as long as you can hear it You must release the button when you stop hearing any sound You will be presented to very weak sounds so it is very important that you stay relaxed and concentrated and that you react as fast as you can At the end of each test part the left or right LED of the pushbutton depending on the tested ear will be turned on Since your task requires a high concentration you should take a break of at least 3 minutes to relax You may leave the room during that time 3 DPOAE measurements A prove will be inserted in your ear canal in a similar way as inserting an ear plug It does not hurt The measurements will be carried out only in the same ear as in the previous test You will hear a two tones sound that will increase in frequency You don t have to do anything Just try to relax not to move and do not swallow The test lasts approximately 20 30 minutes You are very welcome to ask any questions you need Please inform to the experimenter if think you might be under conditions that can affect to your performance during the test i e if you have a cold hangover etc The results of the experiment will be used in a 10 semester project and they will remain anonymous Thank you very much for your collaboration 108 Questionnaire Full name Age Gender O Male Fe
56. ard deviation between subjects 11 3 4 Analysis of individual results From the visual inspection of the DPOAE measurement individual results presented in Appendix C the subjects have been gathered into three differ ent groups The ripple prevalence in the threshold curves see Appendix B within each group of subjects is also analyzed The specific characteristics of each group are detailed next e Group 1 Constituted by subjects 2 and 11 These subjects present a very high DPOAE ripple prevalence in the measured frequency range Furthermore the ripples are in general significantly high more than 10 dB SPL on average When looking at the prevalence of the threshold ripples in Figure it can be seen that it is higher than 70 96 in both cases e Group 2 Constituted by subjects 1 3 5 6 8 9 10 and 12 These subjects present also significant amount of DPOAE ripples but not as high as the prevalence shown by Group 1 The height of the rip ples within the measured frequency range is in general lower than 10 dB SPL The prevalence of the threshold ripples see Figure is around 40 96 for subjects 1 3 5 Subjects 6 and 12 have a prevalence around 50 subject 9 has almost no threshold fine structure and subject 10 presents around 60 96 of threshold ripple prevalence e Group 3 Constituted by subject 4 and 7 These subjects present al most no DPOAE fine structure however their results show the highest 57 CHAPTER 11 R
57. asurement system 25827 Br el amp Kjar 2133 Real time frequency analyzer 08596 B4 105 Table E 1 Equipment used for high resolution audiometer calibration The Sennheiser HDA 200 audiometric headphone was mounted on the B amp K 4153 artificial ear which complies with IEC 60318 with a force of approx 101 APPENDIX E HIGH RESOLUTION AUDIOMETRY CALIBRATION AND NOISE ESTIMATION 10 N according to ISO 389 8 The force was obtained using a 0 5 1 plastic bottle filled with water as a weight E 1 1 Headphone Transfer Function The transfer function of the audiometric headphone was measured with the MLSSA Acoustical Measurement System The output of the computer was connected directly to the Pioneer amplifier To provide the required power for the B amp K 4134 measurement microphone in the artificial ear the B amp K 2807 power supply was connected between the microphone and the input of the computer The MLS signal was recorded with the default setup and a 10 times average It was saved in time and in frequency domain for further analysis E 1 2 Equipment Calibration The equipment was calibrated with a sinusoidal signal at 1 kHz and differ ent gain factors in Matlab The signal s was created using the following equation s y sin27 fot E 1 where fo 1 kHz and g is the gain factor set in Matlab With the B amp K 2636 measuring amplifier the output voltage of the microphone was measured The apparent sound pressure
58. ation about the standard deviation from the average reference for audiometric results using the Sennheiser HDA 200 headphones is not provided in the ISO 389 However in the reference curve for the HDA 200 headphones reported by Poulsen et al Han 89 the stan dard deviation within the studied frequency range is around 5 dB SPL Therefore the average data from the high resolution audiometry will fall into the range of the standard deviation from the average reference curve The average of the determined high resolution thresholds is still higher than the reference This can be expected since almost non of the subjects re ported an absolute threshold higher than 0 dB HL at any frequency The differences between the population that participated in the current experi ment compared to the experiments conducted to obtain the reference curve given in ISO 389 have to be considered For instance the number of sub jects that participated in the current experiment is smaller and they are also older regarding the average age The average values of the threshold obtained with the standard audiometer are higher than the reference and the results provided by the implemented audiometer According to Lydolf Lydo 99 the ascending method produces _as 11 2 High resolution audiometer reliability very high threshold level when compared to the 50 detection level defini tion of the threshold 11 2 High resolution audiometer reliability A
59. audiometer is analyzed as the repeatability of the test retest experiment The aim of this analysis is to check how reproducible the screening of the threshold fine structure is when the test is performed twice by the same subject Thus the similarity between the shape of the thresholds is rated by calcu lating the correlation coefficient r as follows Heis 08 Ty T1 e T5 T PAL me e 10 1 where T and T5 are the vectors containing the two determined thresholds T and T5 are the mean values of the threshold curves and e denotes the dot product of two vectors Since the implemented high resolution audiometer is designed to track threshold fine structure the reproducibility of ripples is also analyzed The difference in the ripple prevalencd among the analyzed frequency range is studied The reliability of position and height of the ripples should be checked by reporting frequency shifting of the extreme values and deviation in ripple height and ripple spacing The ripple prevalence is the percentage of DPOAE or threshold curves classified as fine structure within a specific frequency range 44 10 3 Analysis strategy 10 3 3 Relationship between DPOAE and threshold fine struc ture By simple visual inspection of the shape of threshold and DPOAE curves it is very complicated to try to establish a relationship between them Hence the implemented fine structure detector algorithm described in Chapte
60. aural listening in free field In this case a different reference has to be considered than e g for a standard audiometry where the sound is presented just at one ear each time with headphones For monaural headphone playback the resulting threshold curve shows a higher sound pressure level and a slightly different shape compared to the binaural free field threshold Moor 07 A reference resulting from such experiments especially for audiometric equipment is given in ISO 389 13 CHAPTER 3 THE AUDIBLE RANGE AND HEARING THRESHOLDS 3 2 The audiogram Audiometric measurements are developed to detect individual differences from an average human hearing threshold In order to present those dif ferences a so called audiogram is typically used The individual hearing threshold is normalized to an average reference as proposed in ISO 389 A specific logarithmic scale called hearing level HL is accepted for thresholds specified with this method Hence if a subject shows exactly the same hear ing threshold as the average of healthy humans its audiogram will be 0 dB HL for all frequencies However an increased hearing threshold is typically plotted downwards to illustrate a loss of hearing Moor 07 Figure 3 2 shows a typical hearing loss compared to the reference expressed in sound pressure level and hearing level The threshold in this example is 40 dB higher than normal at 1 kHz 20 Sound Pressure Level dB SPL
61. be observed between the two methods when comparing in dividual results The difference in level between both audiometric measure ments is evidenced For Subject 6 the shapes of the two thresholds are similar However for Subject 2 there are significant differences i e at 1800 46 11 1 Comparison of standard and high resolution audiometry results X Standard threshold High resolution threshold Shifted High resolution threshold Subject 2 Subject 6 15 15 6 Shifted AHL 10 dB 48 Shifted AHL 2 4 dB 5 5 0 0 PS r B 10x 77 git XE a Sew ad S 1 e a E wr DN 15 Keng ga F X TEE 20 25 x 30 i 30 1600 2000 2500 1600 2000 2500 Frequency Hz Frequency Hz Figure 11 1 Comparison of the standard and high resolution audiometry results for subject 2 and 6 Hz From the individual data it can not be concluded why such differences appear Consequently a Two sample t Test is conducted in order to check whether the thresholds determined with the standard and high frequency resolution audiometers present a similar distribution Since the Two sample t Test is conducted in order to check if the mean of two groups of data differs the shifted high resolution threshold values are used for the analysis Seven frequency points common for both measurements are analyzed For each analyzed frequency the distribution of the two groups of 12 threshold values from the 12 s
62. ce across the subjects in 1 3 octave bands for the threshold blue and DPOAE black fine structure as a function of 2 4 fa left fo middle and fi f2 right The errorbars represent the standard deviation between subjects 97 APPENDIX D THRESHOLD AND DPOAE FINE STRUCTURE COMPARISON OF RESULTS D 2 Ripple spacing The average and standard deviation of the threshold and DPOAE ripple spacing calculated for each subject within the frequency range from 1400 Hz to 2800 Hz is presented in Figure 6 7 8 Subjects 10 11 12 Threshold EJ DPOAE as function of fip Hg DPOAE as function of f EE DP OAE as function of 1 1 Average Figure D 15 Average ripple spacing in octaves for each subject The threshold and DPOAE ripple spacing is also calculated in 1 8 octave bands for each subject and averaged over all the subjects The results are shown in Figures and D 17 where the spacing is given in octaves and Hz respectively Ripple spacing octaves 2 Ripple spacing octaves 2 Ripple spacing octaves 2 0 15 0 15 0 15 0 1 0 1 0 1 Au 0 05 0 05 0 05 0 0 0 1000 2000 4000 1000 2000 4000 1000 2000 4000 i 1 2 2 fs Hz fs Hz ft Hz Figure D 16 Average of the ripple spacing across the subjects in octaves in 1 8 octave bands for the threshold blue and DPOAE black fine structure as a function of 2
63. cribed in the following sections 10 3 1 Comparison of standard and high resolution audio metry results The main purpose for conducting the standard audiometry test is to ob tain an estimate of the absolute threshold level at a discrete number of frequencies Since threshold tracking methods are not expected to be pre cise absolute threshold level estimates Heis 08 the standard audiometry performed in a wider range is conducted to provide information regarding the hearing status of the subjects 43 CHAPTER 10 EXPERIMENT DESIGN However the results from the standard and high resolution audiometry tests can be compared to check whether the thresholds present similar distribu tion The common frequency points within the two audiometries can be analyzed for this purpose At each analyzed frequency the results across the subjects from the two audiometries represent two groups of data To check whether the two groups of data are significantly different or not or equivalently how much the mean of the thresholds differs a Two sample t Test can be conducted thus considering the null hypothesis Ross 04 e Ho Hz My the threshold distributions do not present significant differences The average across the subjects at the common frequencies of both tests is compared with the ISO 389 reference curve of the hearing threshold 10 3 2 High resolution audiometer reliability The reliability of the implemented high resolution
64. ction component Prefi can be evaluated as a function of the distortion component Pis with Presi Ra Wap Ra Wap Paist 5 3 where Zy Wapy is the complex relation between the apical and basal traveling wave amplitude of the distortion component The summation of the two sources predicts a DPOAE wave amplitude P 1 a wap Ra wap Past The waves generated by the two sources will also be reflected at the 2 P dB frequency Figure 5 2 Threshold fine structure pattern as a result of 1 1 RaRo with Ra Ra e 2 pa x log f Ral Ro and i are set to be constant for this illustration and R Rp lt 1 base Consequently a more precise description for P is given by Talm 98 Talm 991 Po wap _ 1 Ra wap Zalwap Paist Wap f Ra Wap Ro Wap x 1 Raat RR 5 4 Similar to Rp the variation of Z4 with frequency is expected to be much lower than the phase variation of Ra Hence the variations of Pref Wap Pa PalwWap appear to be dominated by the phase variations of Ra Comparing Equations 5 2 and 5 4 the periodicity in both the threshold the DPOAE fine structure seems to be mainly the result of pa arg Ra Hence the frequency spacing between adjacent fine structure minima is expected to be similar in both cases according to these simplified models Talm 98 In contrast the shape of the different fine structures is also influenced by the absolute values of each r
65. d and how it is used for clinical hearing assessment is explained in this chapter Furthermore the origin of threshold fine structure and its characteristics are described The determination of the absolute hearing threshold is one of the most com mon psychoacoustic measurements It is performed by presenting sinusoidal sounds of different levels to the test subject under the condition that no other external sounds disturb the experiment The absolute hearing threshold is defined as the level at which a sound at a specific frequency is just audible in the absence of any other sounds It is determined experimentally as the level at which the stimulus is detected in 50 96 of repeated trials Lydo 99 The absolute threshold is different from subject to subject but a strong de viation from the average of normal listeners is often an indicator of auditory system damage 3 1 Audible range The human hearing is limited in frequency and sound pressure The area enclosed by those limits is described as the audible range which is illustrated in Figure In the cochlea sound with an appropriate pressure normally produces sensations in a range between 20 Hz and 20 kHz Zwic 90 The absolute hearing threshold defines the lower limit of the sound pressure in the audible range whereas the higher limit is described as the threshold of pain 12 3 1 Audible range All important spectral components of speech and music are within the au
66. d in order to screen the threshold fine structure of the subjects A high frequency resolution audio metric system was implemented for that purpose The system is based on the FINESS algorithm described in Chapter The algorithm was imple mented in MatlaH The calibration of the system and the noise estimation are described in Appendix E B 1 Setup The high resolution audiometry test was performed in the Audiometry Room B4 103 at Aalborg University Fredrik Bajers Vej 7 9220 Aalborg E Wall connections to the neighboring Control Room A B4 105 allow to place the main equipment in separate room The system s setup is shown in Figure The test subjects were seated in the Audiometry Room with the headphones put on They were given pushbutton and asked to press it for as long as they heard a tonal sound and to release it whenever they stopped hearing it The experimenter stayed in the Control Room A to control the test Since the algorithm runs automatically the experimenter had to make sure that the hardware connections were correct the headphones were settled in the right position and the data was saved correctly after the test The experi menter also had to control the necessary time breaks during the test The Matlab files of the implemented high resolution audiometer can be found in the enclosed CD T8 B 2 Equipment Control Room A B4 105 Amplifier Pioneer A 616 Attenuator B4 103 Audiom
67. e not reliable However the purpose of this experiment was to evaluate the height and the periodicity of low level variations of the individual threshold For this analysis the absolute threshold level is of minor importance than the indi vidual shape The reliability tests showed a high correlation between the same experiment performed in different weeks Additionally a comparison with another audiometric method such as the ascending method showed no significant differences This approved the method to be valid for the purpose of screening threshold fine structure Another compromise had to be made regarding the measured frequency range Just one octave was screened with the high resolution audiometer 65 CHAPTER 13 DISCUSSION which lasted approx 20 25 minutes For audiometries a longer time is not recommended since a very high concentration is required Unfortunately the range which allows to analyze both excitation regions of the DPOAE regarding the threshold is therefore very small approx 1 3 octaves The limited time was also a reason for choosing just one set of parameters for the DPOAE measurements A variation of L L or f2 f might have shown different and interesting results The test was performed with 12 subjects 7 male 5 female which were between 20 and 31 years old None of the subjects showed either a severe hearing impairment or an exceptional high sensitivity Since the aim was to find a general
68. ecially on the otoacoustic emissions and hearing threshold fine structure As already mentioned in Section 3 3 reflections along the basilar membrane are involved in the hearing threshold fine structure When the reflectance exceeds 1 those reflections cannot longer be considered as passive This suggests that the cochlea produces an active mechanical response to an acoustic stimulation Kemp 79 Further observations revealed that the time delay between the initiation of a wave and its returning reflection at the base is much higher than expected regarding propagation time Hence the reflections do not occur before a certain reaction time passes Kemp 79 Similar time delays can also be observed when measuring otoacoustic emis sions e g TEOAE see Section 4 2 In fact TEOAE and threshold curves show the same degree of frequency selectivity resulting in coincident min ima and maxima Kemp 79 Furthermore spectral peaks of spontaneous otoacoustic emissions are observed at frequencies of T EOAE and threshold maxima Talm 98 Talmadge et al developed cochlear models to evaluate fine struc tures of threshold and otoacoustic emissions based on the research by Zweig and Shera The variations of the apical reflectance with frequency appear to be the main reason for fine structure see Sections and 4 3 Other involved parameters such as the transformation through the middle ear or the basal reflectance show just little variations with fr
69. ed by the vibration of the eardrum driven by the cochlea Kemp 07 Although this phenomenon is still not completely understood experiments have demon strated that OAE are directly related to the active mechanism of the cochlea see Section 2 3 2 The fact that the cochlea induces the vibration of the eardrum suggests that an imbalance of fluid pressure between the oval and round window must exist This imbalance is only possible if an additional energy of vibration different from the primary hearing process exists and is transmitted back wards to the base of the cochlea causing the middle ear motion The middle ear impedance matching will cause the vibration of the eardrum _16 4 2 OAE measurement techniques It is believed that the additional energy of vibration is generated by the action of the OHCs as a cochlea amplifier see Section P 3 2 The back wards transmission of this energy can be explained by the reverse traveling waves theory However according to Kemp Kemp 07 it has not been ex perimentally demonstrated directly yet that the reverse traveling waves are involved in the generation of OAE 4 2 OAE measurement techniques The sound pressure generated by the eardrum back to the ear canal is very small It is only enhanced at high frequencies by the action of the ear canal as a horn Therefore to record OAE the ear canal has to be sealed thus providing a confined air volume that can be moved by the small v
70. ed that the DPOAE fine structure presents higher ripple prevalence at 2 5 kHz Therefore it is chosen to perform the measurements of the DPOAE fine structure within a frequency range so that both f and the distortion product frequency fap cover one octave band centered at 2 kHz from 1 4 to 2 8 kHz Thus 17 pairs of primary tones are presented within a range of 200 Hz for f lt 3000 Hz and within a range of 400 Hz for f gt 3000 Hz with f from 1300 Hz to 4700 Hz 8 2 ILO96 The DPOAE measurements function of the ILO96 system is used for screen ing the DPOAE fine structure by means of the DP gram test A DP gram test consists of measuring the distortion product 2f f2 amplitude with fixed stimulus intensities L and L and primaries ratio fo f at several discrete frequencies fa The DP gram test allows to setup the following parameters that remain constant during the measurements e level of the primary frequencies L and La e primaries ratio fa fi e frequency resolution CHAPTER 8 DPOAE MEASUREMENTS e f central value of the sweeping In order to measure the DPOAE fine structure a high resolution is needed Therefore to screen the frequency range specified in Section several DP gram test should be carry out due to the resolution limitation of the ILO system The DPOAE measurements procedure is detailed in Appendix Fine structure detection algorithm This chapter introduces an algorithm to class
71. ef clicks Emissions can still be recorded with a hearing loss up to 50 dB Maue 04b Moreover with the DPOAE it is possible to perform hearing assessment at specific frequencies The frequency specificity of the DPOAE measurements makes them suit able when investigating a possible relationship between OAE and hearing threshold fine structure 4 3 DPOAE fine structure Experimental researches and cochlear models by Talmadge et al Talm 98 Talm 99 claim that DPOAE are produced by two sources caused by two different mechanisms One of the sources is placed in the BM at the tonotopic region of primaries overlap near f2 and the other at the quadratic distortion product fap 2fi fe Shaf 03 The first one will be referred to as the distortion component and the second as the reflection component The energy of the distortion component travels both towards the base and the apex When this energy reaches the activity peak of the reflection component in the BM the apical energy is reflected backwards to the ear canal The DPOAE fine 1s 4 3 DPOAE fine structure structure results from the constructive and destructive interference of the two DPOAE sources in the BM Shaf 03 The traveling waves behavior is illustrated in Figure a ae ae base f f 2f f apex PLL SOLO OSOS IIS ING ZI HR 8 LIII IIS IIS UIS UP dist u e Ore fl Figure 4 1 Sketch of the non linear distortion at the region traveling waves of frequencies
72. eflectance and their slow varying phase Hence a direct relation between levels or ripple heights of DPOAE and threshold fine structure is hard to predict since parameters like e g q vary from subject to subject Threshold fine structure screening methods A screening method is necessary to be able to determine the threshold fine structure Two different strategies are analyzed level presentation and fre quency sweeping Finally the choice of method for the implementation of the high resolution audiometer is argued A screening method for detecting threshold fine structure requires a higher frequency resolution than the classical pure tone audiometry Therefore the method should be not only precise regarding the threshold s shape but also fast since high resolution audiometry tests are long and require of a high concentration of the subjects Tracking methods are hence well suited for high resolution audiograms because of their speed l In audiometric tracking methods the stimulus presented to the subject varies on time These variations are usually either in level or in frequency whereas in some cases both frequency and level are varied simultaneously The sub ject is normally asked to press a button for as long as a stimulus is heard The speed of the tracking will influence the method s precision Moreover the tracking strategy i e the choice of constant or varying level and fre quency will affect the results Both track
73. equency c CHAPTER 5 MODELS OF DPOAE AND HEARING THRESHOLD FINE STRUCTURE During an audiometry the ear canal is excited by an external stimulus with the driving pressure Pj resulting in a traveling pressure wave in the cochlea A simple illustration of the cochlear response to an initial pressure wave b w is shown in F igure 5 1 R Ra base tonotopic location apex b I Rab Ra Ry b 2 p Ra Ryb Ra Rib Figure 5 1 Schematic illustration of the reflections due to excitation of the pressure wave b at the stapes Redrawn from Talm 98 The threshold microstructure model suggests a direct relation between the ear canal driving pressure Pay w and the BM displacement amplitude w at the tonotopic location Under the assumption that the transformation of Pay through the middle ear is constant over a small frequency range it can be stated that Eo Palo Paplio b w 5 1 where P w is the resulting apically traveling pressure wave including all apical and basal reflections Ra Ry of b w Talm 98 Palw _ 1 b w E 1 RaRo 1 RaRp RaRo 5 2 Considering that n is the minimum detectable BM amplitude and the phase of Ra varies logarithmically with frequency the threshold level P a w Parlw amp n shows a periodic pattern as illustrated in Figure A similar interpretation can be made for the DPOAE fine structure see Section 4 3 In this case the complex amplitude of the refle
74. equency region of damage the fine structure disappeared the amplitude of the DPOAE recorded was flattened However if the frequency region of impairment cor responded to the distortion component f2 the DPOAE level was reduced but the fine structure could still be observed Therefore not only the level of DPOAE can be used in hearing diagnosis If no fine structure is present a cochlear damage at the reflection component fap can be expected 19 CHAPTER 4 OTOACOUSTIC EMISSIONS Wdp 2n 4n 6n ST qrefl Figure 4 2 Two generation source components phasor diagram left and DPOAE fine structure pattern right where a is the am plitude of the reflection component and Pref is its vary ing phase responsible for the constructive and destructive sources interference and dependent on the distortion product frequency Wap The amplitude and phase of the DPOAE signal are a and Ye respectively The amplitude of the distortion component is adist Models of DPOAE and hearing threshold fine structure The previous chapters revealed that the active nonlinear process in the cochlea is responsible for the generation of otoacoustic emissions and it is involved in the appearance of fine structure in the human hearing threshold Hence a direct relation between both phenomena is presumed In this chapter pos sible similarities and dissimilarities regarding previous studies and cochlear models are presented focusing esp
75. equency track ing procedure based on the level presentation strategy The method presents high frequency resolution and at the same time the precision of the screen ing is ensured by performing several repetitions and consistency checks However the method is very fast since only one reversal per frequency is performed relying on the fact that using small frequency steps the threshold does not vary abruptly from one point to the next The algorithm of the threshold fine structure screening method implemented is described in Chapter 7 27 The FINESS algorithm The high resolution audiometer is implemented using the FINE Structure Screening algorithm which has been developed by Heise et al Heis 08 The procedure and the characteristics of this method are described in this chapter As stated in Section 6 3 the FINE Structure Screening FINESS algorithm is a fixed frequency tracking method for threshold fine structure screening based on the level presentation strategy The algorithm has been implemented in MatlaH The system setup and calibration are described in Appendixes B and Heise et al checked the reliability of the FINESS algorithm with a test retest experiment The results showed a high agreement when subjects performed the audiometric test twice The accuracy of the shape of the threshold determined with the FINESS algorithm was also tested by com parison with an adaptive three alternative force choice 3 AFC
76. etry Room Headphone Sennheiser HDA 200 OO Power Supply E Pushbutton Figure B 1 Setup used for high resolution audiometry test B 2 Equipment The equipment used in the high resolution audiometer setup is detailed in Table Model Type AAU Room no FujitsuSiemens PC with RME DIGI96 8 PST 60907 B4 105 Computers sound card Pioneer A 616 Stereo amplifier 08340 B4 105 Attenuator 40 dB B4 105 B amp O SN16A Power supply 08013 B4 105 Sennheiser HDA 200 Audiometry headphone 33378 B4 103 Pushbutton B4 103 Table B 1 High resolution audiometer setup equipment The headphones and pushbutton in the Audiometry Room were connected to the equipment in the Control Room A through the wall This way the Audiometry Room was completely isolated from the acoustic noise produced by equipment Figure B 2 illustrates the system connections The necessary connections to send a sound signal to the headphones from a Matlab program using the computer in the Control Room A are described in the following e The internal sound card is attached to the PCI slot of the computer e The line output of the sound card is connected to the power amplifier 79 APPENDIX B HIGH RESOLUTION AUDIOMETRY TESTS JOURNAL Control Room A Audiometry Room B4 105 Wall B4 103 Sound Card Stereo Amplifier Attenuator Parallel Paral
77. etry test in the given frequency range is as follows a A 4 Results 1 Select the Auto Threshold softkey 1 program in the menu for special test softkey Special 2 In the Setup change the frequency resolution to 6 points per octave and press Setup again to return to the test program 3 Make sure that the subject feels comfortable in the cabin has the headphones in the right position and understands the task 4 Give the subject the correct pushbutton for their response blue to test the left ear red to the right ear 5 Select the ear to be tested with the key L R Shift left is default 6 With the knobs for Frequency and Level select 1250 Hz and 40 dB HL as a starting point and run the test for this frequency by pressing Sing Frq Start softkey 5 When finished the determined threshold will be marked with an X left or O right 7 Repeat the same procedure with the following frequencies 1250 1400 1600 1800 2000 2240 2500 2800 3150 3550 4000 4500 Hz Start always at 40 dB HL 8 When the test is completed write down the threshold values of the subject A 4 Results The results obtained in the standard audiometry test are shown in Figures A 2 and A3 in a frequency range from 1250 to 4500 Hz and a resolution of 6 point per octave The cross symbols indicate the absolute level in dB HL determined at the correspondent frequency specified in the x axis
78. fine structure analyzed in 1 3 octave band The errorbars represent the standard deviation between subjects 11 3 2 Spacing The ripple spacing is analyzed in order to establish a possible relationship between the threshold and DPOAE fine structure periodicity The threshold and DPOAE ripple spacing in octaves is presented in Figure for each subject within the frequency range from 1400 Hz to 2800 Hz The individual results fluctuate between 1 25 and 1 6 octaves 53 CHAPTER 11 RESULTS 0 2 Threshold 0 18 i E 1 a DPOAE as function of f v 0 16 m ze TE E MN DPOAE as function of f 0 14 0 12 T ses e m 2 E T m 0 08 0 06 0 04 ipple spacing octave 0 02 1 2 3 4 5 6 7 8 9 10 11 12 Average Subjects Figure 11 6 Ripple spacing in octaves for each subject The ripple spacing analyzed in 1 8 octave bands for each subject and aver aged over all the subjects is presented in Figure 11 7 in octaves and in Hz respectively The threshold ripple spacing fluctuates around 1 10 octaves with a maxi mum average spacing over all subjects of 1 8 octaves and a minimum of 1 15 octaves At the lowest frequency the average spacing is 70 Hz and it increases up to 193 Hz on average at higher frequencies In the literature there is no agreement on the exact periodicity range of the threshold fi
79. from the highest level allowed 50 dB HL while in the second run the level tracking begins at 20 dB HL and increases until the button is pressed The consistency between the two determined thresholds is checked An offset difference is permitted since this method focuses on the threshold s shape rather than on the level as stated before The consistency between the two curves is checked at each single frequency taking into account the neighboring values The thresholds are multiplied by a window w of 11 samples centered at the frequency of study The 3The screening of the threshold as described in this section is performed by means of the Matlab function Heise algorithm up down included in the enclosed CD The 11 samples window w is generated by the Matlab function weighting window m included in the enclosed CD 30 7 2 Repetitions and consistency checks windows consist of 2 samples of raised cosine raise 7 flat samples followed by 2 raised cosine fall samples The standard deviation of the normalized difference between the two thresholds windowed data is calculated in each 11 samples segment as follows n n 2 D y Ai Ds Ax 7 1 n 1 where Tinorm fi wif und 7 2 anor fi WP Y weh 7 3 Ai Tinorn fh Tano 0 14 T f and To fi represent the two threshold values at each frequency fi and n 11 is the number of frequency samples in each segment s If 3 exceeds t
80. he button is released Nowadays a modified version with fixed discrete frequencies is often imple mented The level increases or decreases according to the subject s response while the frequency is kept constant After certain number of reversals the methods stops The hearing threshold is estimated by averaging the mid point between the tracked peaks and valleys Lydo 99 Parameters as the attenuation rate the minimum number of reversals and the maximum deviation between peaks and valleys influence the efficiency of this audiometric method This way a more accurate threshold level can be obtained at specific frequencies compared to the original B k sy audiometer 6 2 Frequency sweeping strategy An alternative to the level presentation strategy is the frequency sweeping While the level of the stimulus is kept constant the frequency varies on time In the following this strategy is explained using the Audioscan developed by Meyer Bisch reviewed in Zhao 02 as an example The Audioscan is based on iso hearing level frequency sweeps Its efficiency was tested using the B k sy method as reference Laro 97 The high correlation between both methods confirmed the validity of Audioscan 6 2 1 The Audioscan method The Audioscan method is based on constant level frequency sweeping A constant level stimulus is presented in octave intervals The subjects are 25 CHAPTER 6 THRESHOLD FINE STRUCTURE SCREENING METHODS
81. he value of 2 5 dB the consistency of the screening at that specific frequency is not valid In the frequency regions where the two determined thresholds are not con sistent a third repetition is performed The minimum number of frequency points within a frequency range to be repeated is 11 If the consistency check fails for a frequency range smaller than 11 frequency steps it is extended to the neighboring frequencies If the number of frequency point between two non consistent ranges is less than 11 the ranges are unified In the remeasured frequency range the two more consistent thresholds are determined by finding the minimum X If the threshold obtained in the third run is more consistent than one of the other two the third curve re places the other In the limit frequencies of the repeated range the transition is smoothed by cross fading between the replaced and the third threshold An example of a cross fade between two thresholds is illustrated in Figure The transition points between thresholds are calculated as follows Toy fd 1 a T aly 7 5 where a varies in steps of 0 125 from 0 to 1 or from 1 to 0 depending on the remeasured frequency limit Ty corresponds to the threshold from 5The consistency check between the two thresholds measured is performed by the use of the Matlab function consistency check m included in the enclosed CD 31 CHAPTER 7 THE FINESS ALGORITHM the third repetition of the test in
82. ibration of the eardrum A classification of the OAE can be made depending on the measurement technique and on the type of stimulus applied This classification is de scribed in the following together with the method of analysis used in each case to separate the emissions from the excitation signals Kemp 07 e Spontaneous otoacoustic emissions SOAE No stimulus signal is ap plied In some healthy ears oscillations are produced in the cochlea that feed back itself producing OAE that can be recorded without using any external stimulus Kemp 07 The observation of SOAE is denoted by peaks that stand out from the background noise spectrum The noise spectrum is obtained as the average of the power spectrum of the signal recorded Maue 04b e Stimulus frequency otoacoustic emissions SFOAE are measured by applying a single tone The OAE generated are of the same frequency and they are produced at the same time as the stimulus The non linear characteristics of the SFOAE are used to separate the emission from the stimulus An alternative to observe the SFOAE is to apply a suppressor tone of a different frequency Kemp 07 e Transient evoked otoacoustic emissions TEOAE can be measured by playing a narrow band tone burst stimulus or by broad band short duration stimuli lt 3ms applied repetitively Kemp 07 A delayed low level signal which contains the TEOAE is recorded in the ear canal The TEOAE can be separated from the stimu
83. ify fine structure in DPOAE and high resolution threshold measurements The procedure for the detection of separate ripples is outlined as well as the choice of the parameters for the classification The existence of fine structure in hearing threshold or otoacoustic emis sions is often determined just by visual inspection In order to analyze the obtained data an objective classification of the fine structure is required Hence an algorithm has to be developed to detect fine structure in threshold or otoacoustic emissions and to extract the main characteristics of a ripple These characteristics allow an overall statistical analysis of the experiment results even though the fine structure varies from subject to subject The implemented fine structure detector is based on two different algorithms Heise et al introduced a method for the detection of significant extreme values in threshold fine structure whereas Reuter and Hammershgi developed an algorithm for the classification of DPOAE fine struc ture The determined fine structure is separated into single ripples which are analyzed individually A ripple is characterized by a maximum and two neighboring minima As stated before a more detailed description of the ripples is needed Hence the main characteristics are defined by the follow ing parameters e ripple spacing which describes the frequency spacing between the two minima e ripple height which is the level difference be
84. ing strategies level presentation and frequency sweeping are analyzed in the following supported by an example A discussion of their advantages and disadvantages is also presented at the end of this chapter 6 1 Level presentation strategy To determine the absolute hearing threshold an auditory stimulus is pre sented to the test subject The intensity of the stimulus is varied in or der to find the limit level between an audible and an inaudible stimu lus This is known as the level presentation strategy according to Lydolf 2a 6 2 Frequency sweeping strategy Lydo 99 where the influence of different level presentation strategies in hearing threshold measurements is examined It was concluded from this study that the B k sy is the most efficient of the methods analyzed The B k sy method has also the advantage that it is standardized for clinical applications Hence this method is further described as an example for the level presentation strategy 6 1 1 The B k sy method Originally the B k sy audiometry was performed with continuous sweep frequency tone from 100 Hz to 10 kHz During the sweep the level of the stimulus presented is varied according to the subject s response with an attenuation rate of 2 3 dB s Lydo 99 The subjects are instructed to press a button while the stimulus is heard and to release it otherwise While the button is pressed the level of the stimulus decreases and it increases while t
85. inimum 4 is neglected because of the maximum between 3 and 4 which may represent separate ripple Threshold fine structure subject 12 DPOAE fine structure subject 4 10 5 5 f ANAM a o 0 1 2 fan o 5 20 i i i 10 i i i i i 2200 2400 2600 1900 2000 2100 2200 2300 Frequency Hz Frequency Hz Figure 9 1 Segments of the threshold curve from subject 12 left and DPOAE curve from subject 4 right Minima which fulfill the criteria with respect to the maximum marked with X are indicated by circles The dashed line shows the finally accepted extreme values which characterize the fine structure The maximum under analysis is considered as subsidiary and will therefore be neglected if one of the following cases occur e No left or no right minimum is detected e No pair of left and right minima are separated by frequency spacing between A fmin and A fas e higher maximum appears between the closest left and right minima meaning that the current maximum does not represent an absolute maximum in the analyzed range e The relation of the frequency spacing of the closest detected minima to the ripple height is higher than c ADS meaning that the lower the height of the ripple the narrower it has to be The latter restriction is due to detections in curves as a result of low fluc tuations and not of real existing fine structure This can be observed in the DPOAE measurement of subject
86. instructed to press a button for as long as the stimulus is heard and to release it whenever the stimulus becomes inaudible At first the frequency is swept at a constant level If no response is obtained the level of the stimulus increases for the next sweep When a subject s response is detected throughout a defined frequency range a new stimulus of higher intensity is presented only over the frequency range where no response was recorded The method s procedure is graphically explained in Figure The first sweep is presented at 0 dB HL starting at 1 kHz to the maximum frequency and then from 1 kHz again to the minimum frequency The limits of the frequency range where the stimulus is not heard are represented by Fy and Fy A new sweep at 5 dB HL is therefore presented from Fm towards P first and F later The central frequency Fm is calculated as Fm VFa1 For As a result two new frequency limits Fa2 and Fa are found Hence a new sweep at a higher level than the previous is presented over the new frequency range Frequency Hz 125 250 500 1k 2k 4k 8k 16k 10 10 5 dB step Scan direction 20 30 Next scan level Hearing Level dB HL 40 50 Figure 6 1 The Audioscan sweeping procedure Zhao 02 With this method notches in the absolute threshold can be screened On the other hand the speed of the sweepings will influence the reliability of the experiment La
87. ions on primary levels in normal and impaired ears II Asymerty in L1 L2 space Journal of the Acoustical Society of America Vol 97 No 4 April 1995 113 BIBLIOGRAPHY Zhao 02 F Zhao D Stephens and C Meyer Bisch The Audioscan a high frequency resolution audiometric technique and its clinical applications Clin Otolaryngol Vol 27 No 4 10 October 2002 Zwei 95 G Zweig and C A Shera The origin of periodicity in the spec trum of evoked otoacoustic emissions Journal of the Acoustical Society of America Vol 98 No 4 October 1995 Zwic 90 E Zwicker and H Fastl Psychoacoustics Facts and Models Springer 1st Ed 1990 114
88. isual comparison between the threshold and DPOAE fine structure detected for each subject s results as a function of the distortion product frequency 2 f f2 the second primary frequency f and the geometric mean y f f The ripple prevalence is calculated in 1 3 octave bands for each subject Thus the percentage of ripples present in each band for each subject are represented in the two gray scale bars The upper bar corresponds to the threshold fine structure prevalence and the lower bar indicates the percent age of DPOAE ripples detected However since the generation of both threshold and DPOAE fine structure are the result of complicated phenomena occurring in the cochlea see Chap ters 4 and it is nearly impossible to establish a relationship by simple visual inspection Therefore in the following sections the different fine structure parameters ripple prevalence spacing and height obtained from the experimental re sults are presented as a function of 2f fa fo and y fi fo Mds gp apnydwe pjouseiu 1 o o o o o o N g E o o ive N a oO o o 7 o o o o o o TT mee See Om Das ap epnyjdwe 3voda dS gp apnydwe pjouseiu 1 o o o o ji o Lal N e Das gp epnyjdwe pjouseiu L o T o o o o N a 0 2500 aff i id 1600 2500 IRE 1600
89. ities and dissimilarities between threshold and DPOAE fine structure are described in Chapter 5 with respect to previous studies and theoretical models Different threshold screening methods are analyzed and the method chosen to determine the threshold fine structure is explained respec tively in Chapters 6 and 7 The characteristics and requirements of the system to measure the DPOAE fine structure are argued in Chapter Chapter 9 introduces the implemented fine structure detector for an objective study of the characteristics of DPOAE and threshold fine structure The experiment design and results are presented in Chapters 10 and A deep analysis of the results and comparison with theoretical models is presented in Chapter In Chapter the experimental and analytical findings are discussed Finally the conclusion achieved and the interesting points for future studies are included Chapters 14 and The attached CD includes the results from the experiment and the Matlab files of the implemented high resolution audiometer the fine structure detectors and for the reconstruction of the DPOAE func tions The members of the group would like to thank Miguel Angel Aranda de Toro Rodrigo Ordo ez and Dorte Hammershoi for their interesting feedback and support and to all the subjects that participated voluntarily in the experiment av 1 2 The human auditory system 3 The audible range and hearing thresholds 3 1 Audible range
90. l acts as an open pipe with a resonant frequency of about 4 kHz due to its length of 25 mm to 35 mm The human hearing is thus very sensitive around this frequency and therefore the risk of hearing damage around 4 kHz is higher Fast 07 2 2 Middle ear The malleus incus and stapes are the three small bones present in middle ear known as ossicles see Figure D 1 The malleus is fixed to the eardrum and the stapes footplate is attached to the oval window The middle ear acts as an impedance matching enabling to transform the acoustics pressure variations with weak forces and large displacements to motion of the liquids in the cochlea with large forces and small displacement Fast 07 The impedance matching is achieved mainly by the area of the eardrum and the stapes footplate ratio but in a small scale also by the lever ratio produced by the difference in length between the malleus and the incus Howa 96 Fast 07 Moor 07 2 3 Inner ear In some literature the inner ear is presented as the cochlea which is the auditory organ itself However the inner ear comprises also the balance organ It is of interest for this project to focus here only on the anatomy and functionality of the cochlea where the mechanical energy transferred 3 2 3 Inner ear to the oval window is transformed into neural impulses that are sent to the brain A cross section of the cochlea is shown in Figure The cochlea is sub divided into
91. lel B Data 1 1 Data 1 1 BNC Power supply Figure B 2 High resolution audiometry equipment connections e The output of the power amplifier is connected to an attenuator of 40 dB to reduce the noise generated by the sound card and the ampli fier and to prevent the reproduction of high level sound through the audiometry headphones e The attenuator output connects to the left channel of the audiometry headphones via an XLR connection through the wall e Both left and right input of the headphones are connected to the wall but only the left receives signal depending on the ear tested Since the used headphones are completely symmetric they can also be worn the other way around hence the right ear can be measured with the originally intended left driver of the headphones The connections to supply power to the pushbutton and to read the data from it are detailed next e The pushbutton is connected via a parallel interface IEEE 1284 to the wall in the Audiometry Room A parallel cable connects the wall with the LPT input of the PC e The power supply in the Control Room provides power from 7 to 12 V to the pushbutton in the Audiometry Room via a BNC 50 Q coaxial connection in the wall B 3 Test Procedure The procedure followed to carry out the high resolution audiometry test is described next 80 B 4 Results 1 Prepare the implemented Matlab program for a new subject introduce the string inputs cor
92. less the implementation of cochlear models can be beneficial to analyze the ex perimental results Results The results obtained from the three tests carried out in the experiment are shown in the Appendices A and C respectively In this chapter a detailed analysis of the results is presented The analysis of the results is divided in three parts following the strategy presented in Section 10 3 The first part consists of comparison of the results obtained from the standard and high resolution audiometries The next part examines the reliability of the implemented high resolution audiometer regarding the shape of the threshold and fine structure detected Finally the last part of the analysis focuses on the relation between the threshold and DPOAE fine structure 11 1 Comparison of standard and high resolution audiometry results No severe hearing impairment is observed in any of the standard audio metry results Neither an exceptional good hearing characterizes any of the subjects A comparison between the thresholds determined using the standard au diometer based on the ascending method and the implemented audiome ter based on the FINESS algorithm is carried out Figure 11 1 shows two examples of threshold obtained by the two audiometries To facilitate the visual comparison between both results the high resolution thresholds are vertically shifted minimizing the square difference between thresholds Differences can
93. lotted as a function of f and Wf fo are very similar Further analysis will only present the comparison of the results as a function of 2f fo and fe regarded in the literature as the two sources responsible for DPOAE generation However it is of course a point for discussion whether the distortion product component is generated closer to fo or fif Since the excitation signal used in the DPOAE measurements was chosen with a difference of primary levels higher than 20 dB Li L2 65 45 dB it is considered that the f2 place will be more representative of the distortion component generation site With L gt Lo the f traveling wave is expanded towards the base and its excitation peak is also shifted in the same direction resulting on the frequency shift of the overlap region towards f according to He and Schmiedt l 11 3 1 Prevalence The ripple prevalence is analyzed to objectively quantify the amount of fine structure present within the frequency range under study This allows to check if the presence of fine structure in threshold involves presence of DPOAE fine structure or vice versa Figure shows the average of ripple prevalence in the threshold and in the DPOAE measurements for each subject within the frequency range from 1 4 kHz to 2 8 kHz It is observed that in general the threshold prevalence is lower than the DPOAE prevalence Subjects 2 and 11 present a prevalence of the threshold ripples higher than 70 Thi
94. lus signal by sig nal processing based on the time delay and non linearity of the OAE Knig 98 17 CHAPTER 4 OTOACOUSTIC EMISSIONS e Distortion product otoacoustic emissions DPOAE are the response of the ear canal to a stimulation consisting on two different frequency tones f and f2 with a frequency ratio f2 fi The non linear be havior of the cochlea generates intermodulation products of the two tones where 2f f2 the most prominent of them Measurements of DPOAE over a specific frequency range are carried out by keeping f fy constant while varying f and fz simultaneously Reut 06 At the ear canal a signal containing the stimulus frequencies f and fa to gether with intermodulation frequencies components can be recorded Knig 98 The techniques most commonly used for clinical applications are the TEOAE and DPOAE Knig 98 With a hearing loss of more than 30 dB the presence of T EOAE is in general not detectable Therefore if such emissions are measured it is an indicator of a healthy ear Maue 04b The high sensitivity of TEOAE make them an appropriate measurement to test young ears It is more successful at low frequencies 1 kHz However it is not possible to record any emissions above 5 kHz To test older ears hearing assessment using TEOAE has to be supported by the study of DPOAE ILO 97 In DPOAE measurements more energy is induced since continuous tones are used as stimuli instead of bri
95. male Please fill in the following questionnaire Feel free to ask any questions whenever you consider it necessary 1 Do you have any known hearing disorder Yes No If yes please specify which kind 2 Have you ever had any serious injuries or illness that might have af fected your hearing Yes No If yes please specify which kind 3 Have you ever worked under loud noise conditions Yes No If yes please specify when and for how long 4 Have you ever been exposed to very high sound levels that might have caused a hearing trauma Yes No 5 Have you been exposed to high noise levels within the last 48 hours Yes No If yes could you describe the situation please 6 Are you taking any medicine at the moment Yes No If yes could you specify which kind please 109 APPENDIX G QUESTIONNAIRE 7 How often do you listen to music via headphones almost never sometimes almost daily 8 Have you ever performed a hearing test before If yes please specify when and where Yes No The information gathered here will be treated as strictly confidential I agree with the storage of my data and their use in this 10 semester master project Date Signature 110
96. mented Matlab function Audiometry Gui m is included enclosed CD The D is necessary to identify the subjects results and the ear input denotes the ear under test L Left or R Right The threshold ripples are detected by the use of the Matlab function FI NESS detector m included in the enclosed CD 81 APPENDIX B HIGH RESOLUTION AUDIOMETRY TESTS JOURNAL Threshold fine structure subject 1 dB SPL H 1600 2000 2500 Frequency Hz Threshold fine structure subject 3 dB SPL H E 1600 2000 2500 Frequency Hz Threshold fine structure subject 5 dB SPL 15 20 i E 1 600 2000 2500 Frequency Hz dB SPL dB SPL dB SPL Threshold fine structure subject 2 1600 2000 2500 Frequency Hz Threshold fine structure subject 4 cee NE Ce A 1600 2000 2500 Frequency Hz Threshold fine structure subject 6 E 1600 2000 2500 Frequency Hz Figure B 3 High resolution audiometry results from subjects 1 to 6 _s2 B 4 Results dB SPL dB SPL dB SPL Threshold fine structure subject 7 1600 2000 2500 Frequency Hz Threshold fine structure subject 9 r4 1600 2000 2500 Frequency Hz Threshold fine structure
97. ments seem however to be too high to explain this phenomenon just by the cochlear amplification Other processes are thus expected to be involved which are not regarded in the simplified models There will always be numerous limitations when comparing real measure ments to cochlear models Just considering the high individual variations will already make the computation of a perfect model impossible 6r Conclusion The purpose of this project was to examine a possible relationship between the fine structures of distortion product otoacoustic emissions DPOAE and those of the hearing threshold For this purpose DPOAE and threshold fine structures of different subjects needed to be analyzed An experiment was hence conducted where the hear ing threshold and DPOAE were screened with a high frequency resolution The theory behind both fine structure phenomena and results from previous researches have been studied beforehand to design the measurement proce dure and the strategy for the analysis of the results Numerous choices had to be made regarding the execution and setup of the experiment which also involved certain limitations A high resolution audiometer was implemented in order to screen the hear ing threshold fine structure of each subject It is based on the FINESS algorithm developed by Heise et al The subject s threshold fine structure was determined in the frequency range between 1400 and 2800 Hz The DPOAE fine
98. ments that constitute the human auditory system are represented in Figure 2 1 Pinna Semicircular canals Auditory nerve Cochlea Eustachear canal tympanic window tube Auditory Eardrum Round membrane Figure 2 1 The human auditory system Source http www skidmore edu hfoley images AuditorySystem jpg The outer and middle ear are separated by the eardrum or tympanic mem brane while the oval window constitutes the boundary between the middle and inner ear The acoustic waves are conducted through the outer ear E3 CHAPTER 2 THE HUMAN AUDITORY SYSTEM The sound pressure variations produce mechanical vibration of the eardrum This vibration is transmitted by three small bones in the middle ear the ossicles The forces applied to the oval window by the ossicles set the flu ids inside the cochlea in motion The cochlea is a snail shell shaped organ present in the inner ear In the cochlea the mechanical energy is transformed into neural impulses that carry the acoustical information to the brain via the auditory nerves Soun 07 The characteristics and influence of each part of the human auditory system in the hearing process are described in the following 2 1 Outer ear The outer ear includes the pinna and the ear canal see Figure 2 1 The pinna is the external part of the ear The complex shape of the pinna causes peaks and dips in the sound spectral information The ear cana
99. minima and maxima Hence it is expected that no abrupt changes will occur due to fine structure in the threshold from one point to the next The algorithm focuses on the shape of the threshold fine structure rather than on the absolute threshold level accuracy Since only one reversal per frequency is presented the determined threshold is expected to be higher than the true threshold when the level of the stimuli presented increases On the other hand the determined threshold will be lower when the level of the stimuli presented decreases This is due to the reaction time of the subjects Therefore the threshold is described by the average between those peaks and valleys The best result to calculate the threshold estimate is obtained by applying locally weighted quadratic regressiox with a span of 9 data points Heis 08 Figure shows an example of a screened threshold using the FINESS algorithm A familiarization period is presented at the beginning of the screening to help the subjects to get accustomed to the method The first frequency of the screened range is kept constant and several reversals are presented before a stable level is reached Therefore the familiarization is implemented to avoid a bias in the threshold screening produced by the period that the sub jects need to get accustomed to the method Once the difference between consecutive detections is less than 1 3 dB it is considered that the value of the determined threshold i
100. n Automatic screening and detection of threshold fine structure International Journal of Audiology 2008 111 BIBLIOGRAPHY Howa 96 ILO 97 Kemp 07 Kemp 79 Knig 98 Laro 97 Lydo 99 Maue 04a Maue 04b Maue 99 Moor 03 D M Howard and J Angus Acoustics and Psychoacoustics Focal Press 1st Ed 1996 ILO OAE Instrument User Manual Otodynamics Ltd Octo ber 1997 D T Kemp Otoacoustic Emissions Concepts and Origins Active Processes and Otoacoustic Emissions in Hearing Vol 30 No Chapter 1 December 2007 D Kemp The evoked cochlear mechanical response and the auditory microstructure evidence for a new element in cochlear mechanics Scandinavian Audiology Supplementum Vol 9 1979 R Knight and D Kemp Relationships between DPOAE and TEOAE amplitude and phase characteristics Auditory Bio physics Group ILO University College London RNTNE Hos pital 330 332 Gray s Inn Road London WC1X SEE United Kingdom December 1998 C Laroche and R H tu A study of the reliability of Au tomatic Audiometry by the Frequency Scanning Method AU DIOSCAN Audiology Vol 36 No 1 18 1997 M Lydolf The threshold of hearing and contours of equal loud ness a study on measuring methods and normal hearing De partment of Acoustics Aalborg university Denmark ph d dis sertation Ed 1999 M Mauermann G Long and B Kollmeier Fine
101. n the results from the measure ments are presented as well bottom _62 12 1 Analytical models The performance of the simulations varied a lot among the subjects Hence 1t was checked if a general trend regarding the groups developed in Chapter 11 could be observed The two subjects of Group 1 showed generally pronounced fine structure in DPOAE as well as in the hearing threshold Since the notches in the DPOAE are very deep it is expected according to the models that the distortion component and the reflection component have similar levels The best comparable results where obtained by setting Rao in the models slightly higher than 1 The pronounced ripples in the modeled threshold are sup ported by comparably high values of Rp Figure 12 3 shows a representative example for the comparison of the simulations and measurements of this group o o 3 oO 5 E I E 2 2 2 i a 2 a E 5 E 5 G D 6 2 lt E lt 2 2 E 2 E E E a E 1600 2500 1600 2500 2i f 20 20 20 20 T T 10 107 z 10 107 Q 2 o 0 0 Sov 0 0 8 o d Y E 2 23 2 2 a E 10 10 8 E 10 10 8 2 22 E 9 89 3 2 20 20 5 20 2 30 30 30 30 1600 2500 1600 2500 Figure 12 3 Comparison of modeled top and measured bottom fine structures of Subject 2 Group 1 Blue lines indicate the threshold and black lines DPOAE A completely different picture of fine structure can be seen for
102. ne structure It varies from one study to another Heis 08 When analyzing the DPOAE ripple spacing as a function of 2 4 fo the average spacing is 1 11 octaves with a standard deviation of 1 72 octaves Hence the spacing increases from 47 Hz at the lowest frequency up to 154 Hz at high frequencies If the DPOAE ripple spacing is now analyzed as a function of f the average spacing is also 1 11 octaves with a standard deviation of 1 69 octaves This corresponds to a spacing from 68 Hz at the lowest frequency to 255 Hz at high frequencies The ripple spacing is however dependent on the parameters defined by the fine structure detector see Chapter 9 When an objective fine structure detector algorithm is designed it is necessary to define some parameters that establish the rules to neglect or accept ripples as fine structure Those parameters are based in general on a compromise between frequency spacing and level height of the ripples meaning that specific values of those are already expected This will itself influence the analysis and could therefore be a reason for disagreement among different studies Nevertheless a similar growth is observed in the DPOAE ripple spacing as a _54 11 3 Relationship between DPOAE and threshold fine structure Ripple spacing octaves 0 15 0 1 0 05 0 n rn n ni n i 1000 1250 1600 2000 2500 3150 4000 5000 f f Hz Ripple spacing Hz 350 300 250 200
103. nt a rj gt Aqist e the amplitude of the reflection component is lower than the amplitude of the distortion component Gre fl lt adist The variations of DPOAE according to changes in the cochlear behaviour are presented in Figure 12 1 The condition of the basilar membrane is described by a simple function E w lt 1 simulating an impaired region in which R and aqist are reduced by up to 30 dB R E w Rao 12 1 adit E w aaisto 12 2 where Rao and aqist are the initial absolute values of Ra and aqist 59 CHAPTER 12 COCHLEAR MODEL SIMULATIONS Amplitude Frequency YX A Amplitude Frequency Figure 12 1 Simulated DPOAE black and threshold blue fine struc ture as a result of an artificial generated excitation function E w The DPOAE fine structure appears to be directly dependent on 2f f2 The upper figure a presents the behaviour with Rao 0 8 whereas the curves for Rao 3 are presented in the bottom b The amplitudes of the DPOAE and the threshold are calculated according to the simplified models presented in Chapter 5 as follows 14 Ba E wap Rave a 1 R E wap Rape Han 1 EE E w Rave 12 3 appoaE E w2 agist o QThreshold E w 12 4 Figure 12 1p illustrates the case where aref lt adist represented by a maxi mum apical reflectance Ray 0 8 for simplicity Z4 1 The DPOAE fine structure is reduced
104. o change their shape length and stiffness The OHCs actively influence the response of the BM to sound When a point in the BM is excited by 10 2 3 Inner ear 1600 Hz 6400 Hz Pell displacement distance from oval window Figure 2 4 Schematic excitation pattern of the BM places using an ex citing signal composed by three separated tonal frequencies 400 1600 and 6400 Hz For the 400 Hz tone the instanta neous traveling waves are also represented The dashed lines represent the envelope which maximum is produced at the tonotopic place in the BM Redrawn from Page 29 a traveling wave maximum the movements of the OHCs increase At low levels the OHCs act as an amplifier increasing the displacement of the BM hence the inner hair cells IHCs are stimulated Moor 07 The IHCs positioned at the BM places that are in motion produce neural impulses Those impulses are carried to the brain by the afferent fibers From the brain the neural impulses travel back to the ear being carried by a different fibers known as the efferent fibers Most efferent fibers are connected to the OHCs It is believed that the active mechanism of the OHCs might be controlled by the efferent fibers Moor 07 The audible range and hearing thresholds The absolute hearing threshold is the lower limit of audible sound pressures and an important element to describe the status of the individual hearing In which way the threshold is define
105. ptimal When the best possible probe fit is achieved press OK to continue the measurements e The stimuli levels are set automatically according to the checkfit data Press OK to start the DP gram measurements e The DP gram measurement starts automatically The measure ments at the 17 points within the current range are repeated continuously Press Normal Stop to finish the measurements when an acceptable S N ratio is achieved and DPOAE amplitude points are above the noise measurements if possible e Press Quick save to save the measurements Input the subjects ID and press F10 and OK to confirm the storage e Press Exit to start a new measurement or to finish the test 5 The measurements are to be saved as spreadsheets once the experi ment is over Each measurement should be loaded pressing Menu Analysis Numerical analysis By pressing any key a pop up menu appears on the screen Select the option save as spreadsheet and save the file The 13 measurements should be loaded and saved as spreadsheets 86 C 4 Results 6 Concatenate the 13 measurements results using the Matlab function dpoae_fine_str C 4 Results Figures C 2 and C 3 show the DPOAE measured in 12 ears from 12 different subjects as a function of the primary frequency f2 The curves of the stan dard deviations from the background noise representing the 95 96 confidence interval are also shown The DPOAE fine struc
106. r 9 is used to extract the characteristic features of the fine structure Thus an objective criterion is obtained providing a suitable tool for comparing threshold and DPOAE fine structure From the detected ripples objective parameters such as the ripple prevalence spacing and height can be analyzed Since the generation of both fine structures is a complex phenomenon and they vary from subject to subject not only average result have to be con sidered but also an individual analysis should be included The frequency range in the DPOAE measurements is chosen allowing direct comparison of the threshold curve with the DP gram as a function of 2 f fa and f for the same subject The tonotopic locations of 2 4 fo and f are regarded as the DPOAE generation sites An individual analysis of the results is made in order to investigate the influence of cochlear mechanisms on the generation of threshold and DPOAE fine structure The comparison of both curves as a function of 2f f2 and f may reveal the influence of the generation sources on the DPOAE fine structure and level The influence of the generation sources is assumed to be dependent on the information given by the threshold regarding the cochlea status Different authors have developed analytical models of the cochlear behavior This models try to explain the complicated generation processes of threshold and OAE fine structure which are still not fully understood Neverthe
107. respondent to the subject s ID and tested ear in the function Audiometry Gui ID ear 2 Make an equipment and connections check according to the graphical guide e Headphones power amplifier is on e Pushbutton power supply is on e Headphones and pushbutton in the Audiometry Room are con nected to the wall 3 Make sure that the subject feels comfortable has the headphones in the right position the left headphone is always on the ear under test and understands the task 4 Leave the subject alone in the Audiometry Room close the door and press start 5 After each part the currently determined curve will be displayed 6 Control the breaks between the test parts at least 3 minutes breaks 7 Check that the data has been correctly saved in a mat file with the subjects ID in the folder hr results B 4 Results Figure and show the results from the high resolution audiometry test The blue solid line represents the threshold values determined at high frequency resolution 100 points per octave The threshold fine structure detector see Chapter lo is applied to the result The detected ripples are indicated by the gray rectangles The criteria used to classify a ripple as fine structure are determined by a minimum level difference between two adjacent extreme values ALmin 2 dB a minimum frequency spacing of A fmin 1 25 octaves and a maximum frequency spacing of A fmax 1 5 octaves The imple
108. ro 97 6 3 Level tracking vs frequency sweeping Since an audiometry is a psychoacoustic test it depends on the responses of the subjects Hence a lack of precision is always produced due to the reaction time that takes the subjects to realize that a stimulus is being heard until the button is pressed These uncertainties will appear on the 26 6 3 Level tracking vs frequency sweeping frequency axis with the frequency sweeping strategy and on the level axis if the level presentation strategy is used For the same measurement time frequency sweeping methods generally pro vide a higher frequency resolution than level tracking methods However the duration of the Audioscan increases with the number of level variations or ripples to be detected and is therefore heavily dependent on the threshold s shape The level resolution of 5 dB used in the Audioscan is not sufficient to detect threshold ripple which can have heights from 2 dB Heis 08 For a comparison with OAE measurements it is of interest that threshold fine structure screening methods are precise in frequency Therefore a fixed frequency level tracking procedure seems more suitable It has to provide a frequency resolution of at least 50 points per octave since periodicities of 1 21 octaves have been observed in previous studies Heis 08 The method named FINESS FINE Structure Screening developed by Heise et al is chosen for implementation It is a fixed fr
109. s corresponds in both cases to a presence of DPOAE ripples of 90 or higher as a function of f This suggests that in general more fine structure is present in the DPOAE than in the threshold curves Just subject 7 presents a higher prevalence in threshold than DPOAE The average across subjects of the ripple prevalence is shown in Figure 11 5 in 1 3 octave bands It can be seen that the prevalence of DPOAE ripples as a function of fo presents the highest value at the mid frequency range centered at 2 5 kHz Similar results are reported by Reuter and Hammershgi Reut 06 The maximum ripple prevalence in the threshold fine structure also appears in the 2 5 kHz band A correlation between the prevalence curves of the threshold and DPOAE as a function of fa can be noticed 52 11 3 Relationship between DPOAE and threshold fine structure Threshold DPOAE as function of a Hg DPOAE as function of f Ripple prevalence 6 12 Average Subjects Figure 11 4 Average ripple prevalence for each subject within the fre quency range from 1 4 kHz to 2 8 kHz Ripple prevalence Ripple prevalence 100 80 60 40 20 0 i i i i i 0 i i i i i 1000 1250 1600 2000 2500 3150 4000 5000 1000 1250 1600 2000 2500 3150 4000 5000 f f Hz f Hz Figure 11 5 Average ripple prevalence across subjects of the threshold blue and DPOAE black
110. s stable A maximum of 16 reversals is pre sented This number is enough to ensure that the threshold of the subjects is determined within 2 dB of their asymptotic threshold Heis 08 The frequency range for the threshold fine structure screening is limited due to time considerations A range of one octave centered at 2 kHz is chosen The high prevalence of DPOAE fine structure in the mid frequency range influences the choice of the range for the threshold screening see Section for further information with this regard Thus the threshold fine The smoothing by locally weighted quadratic regression is performed by using the LOESS method of the smooth m Matlab function For further information see Clev 79 29 CHAPTER 7 THE FINESS ALGORITHM Original Th i j Smoothed Th i pe it I Threshold dB SPL 15 1 1 1 1 1 1400 1600 1800 2000 2200 2400 2600 2800 Frequency Hz Figure 7 1 The threshold obtained by the FINESS algorithm dashed line is smoothed by applying locally weighted quadratic re gression to calculate the threshold estimate solid line with a span of 9 data points structure will be tracked in one octave in the range from 1 4 kHz to 2 8 kHZ 7 2 Repetitions and consistency checks The screening is performed at least twice to ensure the reliability of the experiment To avoid rhythmic responses in the first run the level of the tone starts decreasing
111. structure of hearing threshold and loudness perception Journal of the Acoustical Society of America Vol 116 No 2 August 2004 M Mauermann Fine structure distorsion product otoacous tic emissions and auditory perception Ph D Thesis Insti tut f r Physik an der Fakultat f r Mathematik und Naturwis senschaften der Carl von Ossietzky Universitat Oldenburg 2004 M Mauermann S Uppenkamp P W J van Hengel and B Kollmeier Evidence for the distortion product frequency place as a source of distortion product otoacoustic emission DPOAE fine structure in humans II Fine structure for differ ent shapes of cochlear hearing loss Journal of the Acoustical Society of America Vol 106 No 6 December 1999 B C J Moore An introduction to the psychology of hearing Academic Press 5th Ed 2003 112 BIBLIOGRAPHY Moor 07 Reut 05 Reut 06 Reut 07 Ross 04 Shaf 03 Soun 07 Talm 98 Talm 99 Whit 95a Whit 95b B C J Moore Cochlea Hearing Loss Physiological Psycho logical and technical Issues John Wiley amp Sons ltd 2nd Ed 2007 K Reuter and D Hammershgi Characterization of DPOAE fine structure Forum Acusticum Budapest 2005 K Reuter and D Hammershgi Distorsion product otoacoustic emissions fine structure analysis of 50 normal hearing humans Journal of the Acoustical Society of America Vol 1 No 120 July 2006 K Reuter and D
112. structure was measured with the Otodynamics ILO96 measurement system with the primary levels L1 L2 65 45 dB SPL and the frequency ratio f2 f 1 22 The frequency range for fa between 1300 and 4700 Hz was chosen in order to evaluate the influences of the two sources considered to be responsible for the DPOAE fine structure Twelve subjects participated in the experiment which consisted of three major tests First a standard audiometry was carried out to check the overall status of the subject s hearing in the complete analyzed frequency range The threshold fine structure was screened with the implemented high resolution audiometer in the second test Finally the fine structure of the DPOAE was measured 68 A comparison of the data obtained from the standard and the high resolution audiometer revealed no significant differences in the data distribution across subjects Additionally the reliability of the high resolution audiometer was checked with four subjects in a test retest experiment The two curves obtained for each subject showed a high correlation and thus approved the precision of the implemented audiometer In order to examine a relationship between both fine structures a detection algorithm was implemented to extract individual ripples and specify their characteristics This allowed an analysis of the individual results but also an overall comparison of both fine structures regarding the average across the subjects
113. tion 10 30 25r Threshold dB SPL mean threshold J Prag cross faded curves 10 I T 1 1 1 1 1400 1600 1800 2000 2200 2400 2600 2800 Figure 7 3 Frequency Hz Hearing threshold of a subject determined using the imple mented FINESS algorithm The consistency of the two first repetitions is calculated as the normalized difference between the two thresholds X which is compared with the criteria value of 2 5 dB up A third repetition is carried out in the inconsistent frequency ranges marked by an x The ranges for third screening are extended to a minimum of 11 sample frequencies The thresholds obtained from the three repeti tions are shown middle The two cross faded more consis tent threshold are averaged obtaining the final threshold values down DPOAE measurements A description of the system used to measure the DPOAE fine structure and its specific requirements is presented in this chapter The DPOAE fine structure is measured by using the commercial ILO96 system from Otodynamics The parameters of the system should be setup in order to fulfill the system requirements 8 1 System requirements The amplitude and presence of fine structure on the measured DPOAE depends in the measurement parameters Thus the choice for the ratio of primary frequencies f5 f as well as of the level of the primaries L and La needs to be discussed The prevalence of the DPOAE fine
114. ture detector see Chapter 9 is applied to the re sulte The detected ripples are indicated by the gray rectangles The criteria followed to classify the DPOAE ripples as fine structure are given by the next parameters e minimum level difference between two adjacent extreme values ALmin 3 dB e minimum frequency spacing A fmin 1 25 octaves e maximum frequency spacing A fmar 1 5 octaves The Matlab function dpoae fine str m is included in the enclosed CD The DPOAE ripples are detected by the use of the Matlab function OA E fs detector m included in the enclosed CD 87 APPENDIX C DPOAE TESTS JOURNAL FINE STRUCTURE MEASUREMENTS Subject 1 15 sl n Oo ea me 30 1600 2000 2500 3150 4000 f 2 Subject 3 15 10 5 f n Oo ea Eo 30 1600 2000 2500 3150 4000 f 2 Subject 5 15 10 ad n a ea c 1600 2000 4000 dB SPL dB SPL dB SPL 15 10 10 a Subject 2 1600 2000 2500 f 2 Subject 4 3150 4000 NE 1600 2000 Subject 6 Vinh PY j MINIMA or E il 1600 2000 2500 fo 3150
115. tween the maximum and the average of the two minima 37 CHAPTER 9 FINE STRUCTURE DETECTION ALGORITHM In order to be accepted as fine structure adjacent extreme values have to fulfill specific requirements the level difference between adjacent minima and maxima should be higher than a minimum level AL in and the ripple spacing should be in a range between A fmin and A fmaz Reuter and Hammershgi reported a ripple spacing between 1 21 and 1 6 octaves for DPOAE fine structure whereas Heise et al proposed a criterion for the spacing of two adjacent extreme values mini mum maximum between 1 50 and 1 10 octaves The chosen parameters A fmin 1 25 octaves and Afmaz 1 5 octaves are based on Heise s pro posal and agree with the observations of Reuter and Hammershgi For DPOAE measurements A Lmin 3 dB was chosen according to Reuter and Hammersh i Reut 06 since a high number of low level variations ap pears in the measurements However thresholds show in general a lower ripple height but also less fluctuations which are not considered as fine structure Hence the minimum level criterion applied for the detection of threshold fine structure is set to ALmin 2 dB 9 1 Detection procedure The detection of relevant extreme values which characterize a ripple follows an iterative process Each detection starts at a maximum with the frequency Fmax All minima within the range fmax A fmax fmax Afmax which ful fill the
116. ubjects is compared The p value obtained for each frequency are presented in Table Frequency 1400 1600 1800 2000 2240 2500 2800 p value 0 6909 0 1246 0 3896 0 6487 0 0481 0 0342 0 9814 Table 11 1 p value obtained from the Two sample t Test for the analy sis of the similarities in the distribution of the standard and high frequency resolution audiometry results at 7 common frequencies For a level of significance of 1 96 the null hypothesis can not be rejected for all the frequencies Therefore it can not be stated that significant differences between the distribution of the two audiometric measurements exits The average across the 12 subjects at each common frequency and the cor respondent standard deviation between subjects are shown in Figure 11 2 ar CHAPTER 11 RESULTS The results are compared with the average reference curve proposed in the ISO 389 Sound pressure level dB SPL 30 a i l 1250 1600 2000 2500 Frequency Hz Figure 11 2 Average and standard deviation at the common frequencies of the thresholds determined with the standard audiometer based on the ascending method blue and with the high res olution audiometer black The results are compared with the hearing threshold reference curve proposed in ISO 389 dashed The average curve obtained for the high resolution audiometer is more similar to reference curve Inform
117. ulus intensity the active mechanism and hence the reflectivity would decrease avoiding an excessive growth of the standing wave oscillations This explains why equal loudness contours show a reduced fine structure at higher sound pressure levels Kemp 79 The presence of minima and maxima at specific frequencies highly depends on each individual ear Hence a specific apical reflectance can not be de termined for each frequency or level it rather appears that it changes indi vidually in a random way Kemp 79 Otoacoustic emissions In this chapter the origin of otoacoustic emissions OAE is explained A classification of the OAE depending on the measurement conditions is also presented The arguments for choosing the DPOAE for a comparison with the hearing threshold is given Finally the main theories regarding the gen eration of the DPOAE fine structure are analyzed The OAE measurements are employed as a physiological test for hearing assessment Since they do not require of the active participation of the subject they can be useful when the subject is not capable to perform psychoacoustical tests such as the audiometry method see Chapter B3 i e neonates or young children OAE can be defined as sound generated by the ear Kemp 07 The basic concepts regarding the generation of OAE are presented in the following 4 1 Physiology behind the generation of OAE OAE are sound pressure variations in the ear canal generat
118. ween the gain g and the sound pressure p at the headphones at 1 kHz can be obtained g 1 kHz 3 8 E 2 In order to obtain the gain factor at a specific frequency f different than 1 kHz the normalized transfer function of the headphone see Section E 2 2 is inverted and multiplied in linear scale H 1000 Hz g f g 1kHz Hf E 3 For security reasons this correction is limited to a maximum amplification of 12 dB which might happen for frequencies higher than 12 kHz Audiometric measurements in this frequency range are not recommended with the given setup The gain factor calculated in Equation E 3 is used to generate the pure tone stimulus s of the desired hearing level in Matlal according to following equation compare to Equation E 1 s g f sin2n ft E 4 The implemented Matlab function get_gain m to obtain the gain g for a given hearing level and frequency is included in the enclosed CD 106 Listening experiment instructions The listening experiment you are participating in consists of three single tests for hearing assessment Only one ear will be tested and it will be the same in the three tests The tests will take place in the Audiometry Room B4 103 There will be short time breaks 5 minutes between tests During the breaks you can relax and enjoy the coffee tea and cookies we offer you Test Test 1 Pure tone audiometry This test will take approximately 10 minutes You
119. will be wearing headphones during the test Make sure they fit your ears correctly The red earphone should be placed on the right ear and the blue one on the left ear You also will be given a pushbutton Your task You will hear different tones at different frequencies and levels Since tones around your threshold of hearing will be presented you will need to be very concentrated to detect the lowest tones you are able to hear Relax try to control your breath and press the pushbutton every time you hear a tone Do not try to guess 2 High resolution audiometry The test will last approx 20 25 minutes The test consists of two parts of approx 6 minutes each In each part the test is running continuously without any break so you should try to stay focused during the whole period It is important that you are aware that there are no wrong responses just try to perform in the best way you can Depending on the previous results a third part will be required which will be of equal duration or shorter than the two previous You will be wearing headphones during the test Make sure they fit your ears correctly and the headphone s wire corresponds to the ear under test 107 APPENDIX F LISTENING EXPERIMENT INSTRUCTIONS Test You will find a LED on indicating the button that you have to press Press the button whenever you feel ready to start and the LED will turn off indicating that the experiment begins Your task You will he
120. with no remarkable characteristics Variations among the fine structure e g reduction of ripple height may have different reasons for instance R can be either higher or smaller than 1 Hence the simulations give a lot of different results and are strongly dependent on the choice of the desired characteristics A specific trend for the cochlear reflectances of all the subjects in this group can therefore not be observed 64 Discussion This chapter evaluates the limitations and compromises of the conducted experiment Furthermore the results and their validity are discussed and interesting findings are emphasized Measurements of fine structures in hearing threshold and otoacoustic emis sions require a high frequency resolution and precision regarding small level changes An increasing resolution usually leads to a higher measurement time which is a drawback especially regarding psychoacoustic measure ments Audiometries require full concentration of the subject during the whole test hence it is necessary to use a method which provides both a low measure ment time and a high resolution The implemented audiometer based on the FINESS algorithm provides both requirements but also involves certain compromises One compromise is the absolute threshold level Since the FINESS algo rithm is a tracking method the absolute level might be influenced by several parameters such as the reaction time of the subject and is therefor
121. y dependent on the choices of measurement parameters It has been reported in the literature that the choice of the primary frequency ratio and levels has a significant influence on the DPOAE level and the presence of fine structure The choices made can even produce a shifting of the fine structure minima and maxima along the frequency axis in a DP gram This constitutes an obvious limitation to use DPOAE as an objective audiometry and therefore an appropriate line of research Similarities between the threshold and DPOAE fine structure periodicity are observed in the experimental results thus supporting the analytical models However the models present certain restrictions to explain the generation mechanisms responsible of the threshold and DPOAE fine structure Fur ther investigation of the different parameters involved in the simulation of the cochlear behavior could lead to a more realistic relation between mea surements and cochlear models 70 Moreover the effect of variability between subjects could be reduced by classifying the tested population regarding their hearing status This would facilitate the comparison of DPOAE and threshold fine structure since it might show common characteristics within different groups Finally it could be an interesting future work to analyze the relationship between threshold and DPOAE fine structure by the use of Input Output functions since they appear to be a good estimator of absolute

Relation between fine structures in hearing thresholds and distortion

Contents

Download Pdf Manuals

Related Search

Related Contents