Here - Phonetik
Contents
1. below 10 provide strong smoothing while higher values above 25 leave the contour largely unaffected see Figure 6 Figure 6 A single utterance plotted using manually corrected f0 points silver speckles and two smoothed contours obtained by setting both SMOOTH and SMOOTH to 20 black line and 5 red line respectively Two smoothing values can be specified one for smoothing the manually corrected fO points SMOOTH1 and one for smoothing the contour again after interpolation SMOOTH2 Of course the right choice of smoothing values depends on the characteristics of the speech material and on the purpose of the analysis For intonation research I have often found that it is good to keep SMOOTH2 at 15 and set SMOOTHI to low values e g 5 when eyeballing the data in order to retain only the information about the number and the location of the main pitch events and to high values e g 20 when extracting phonetic features such as the alignment and the scaling of fO peaks 3 How do I acknowledge using mausmooth mausmooth has been developed by Francesco Cangemi at fL Phonetik University of Cologne integrating useful comments from Bastian Auris Anna Bruggeman Samuel Ebner Henrik Niemann Timo R ttger and Simon Wehrle You can acknowledge its use by adding in your paper something along these lines fO contours were extracted corrected and plotted in Praat Boersma amp Weenink 2015 using a customized version of2. real silenceThr 0 03 real voicingThr 0 45 real octave 0 01 real octavejump 0 35 real voiceunvoiced 0 14 endform Erase all Select inner viewport 0 5 2 5 0 5 1 5 Solid line Line width 1 open files Create Strings as file list pitchlist directory Pitch Create Strings as file list list directory wav Sort nfile Get number of strings for i to nfile name Get string i basename name wav Read from file directory name ELABORATION 1 EXTRACT original f track To Pitch ac timestep pitchmin maxcandidates no silenceThr voicingThr octave octavejump voiceunvoiced pitchmax manual correction determine if there is already a manually corrected pitch file exists 0 selectObject Strings pitchlist npitch Get number of strings for k to npitch pitchname Get string k pitchbasename pitchname Pitch if pitchbasename basename exists 1 endif endfor if it exists then open else provide manual correction and save if exists 1 Read from file directory basename Pitch elsif exists selectObject Sound basename View amp Edit selectObject Pitch basename Copy basename View amp Edit pause Confirm Save as text file directory basename Pitch endif Rename manual smooth track Smooth smooth1 Rename smooth Interpolate Smooth smooth2 Save as text file directory basename smooth Down to PitchTier Save as headerless sp
3. visualizes the extracted fO values gives you the time to make manual corrections and proceeds with smoothing and interpolation Of course you don t have to provide manual corrections for all files every single time you run the script Once provided by the user manual corrections are stored in individual pitch files with the name of the corresponding sound file mausmooth only asks for a manual verification when no such pitch files are already provided This means that you will only be asked once to verify each of your files The script opens a form asking basically three things see Figure 3 e the path to both input and output files the working DIRECTORY e the plot parameters the name of the output figure FIGNAME the vertical range of the plot PLOTMIN and PLOTMAX and crucially which combination of RAW MANUALIy corrected and SMOOTHed contours should be plotted For example Figure has been created by ticking both raw and smooth boxes while Figure 2 has raw contours in the left panel and smoothed contours in the right one The advanced user can further customize the plots e g line thickness speckles colour box size by modifying the script directly e the pitch extraction parameters essentially the ones discussed in the Praat manual for the To Pitch ac command with the addition of two values for smoothing SMOOTHI and SMOOTH2 which will be discussed later Run script Input parameters path with final slash direc
4. Figure 4 to resume the script note that the Pause window might end up below the Pitch window The manual corrections are stored in a pitch file then the contour is smoothed interpolated and smoothed again using the parameters provided in the form and stored in a smooth file The smoothed contour is then converted to a HEaderless SPreadsheet file hesp a simple table format which can be used in other environments e g R or Matlab Then the contour is plotted in Praat according to the plot parameters Once the contour is plotted the script processes the following sound file in the path folder in the same way When all sound files are processed the script saves the plot with the filename provided in the input form 2 2 Correction and smoothing choices In order to understand how the script works when using mausmooth for the first times it is useful to plot individual sound files by ticking all three boxes RAW MANUAL and SMOOTH in the plot settings Consider Figure 5 e RAW the fO values extracted by Praat are plotted as red dots as in Figure 1 e MANUAL on top of these the outcome of the manual correction is plotted with silver dots This has the effect of highlighting only those points which were extracted by Praat but were rejected in the manual correction and which now surface as red e SMOOTH the smoothed contour which uses the manual correction as input is overlaid as a black line Figure 5 A single utterance plotte
5. d using all three styles Of course deciding whether a given f0 point extracted by Praat should be retained or discarded is no hard science In order to take principled choices at least a basic knowledge of phonetics and signal processing is required but manual correction is in the end a subjective matter Some readers for example might find that another couple of points could have been rejected in the example in Figure 5 perhaps during the course of the first rise As always decision criteria should be clearly stated and consistently applied In my own practice for example I tend to manipulate reject or change only pitch points during non modal phonation e g the halved values in the second half of Figure 1 or at segmental discontinuities In this latter case I am more comfortable removing isolated points e g the two points in the final flat stretch in Figure 5 which thus surface as red than points in an environment of similar values e g the three or four points in the first rise fall movement in Figure 5 which are not removed and thus surface as grey Once the mechanics of the script are clear and sufficient familiarity with manual corrections is acquired the script can be used to plot multiple smoothed contours only i e by ticking only the SMOOTH box in the plot settings This choice often yields the most readable display of data see Figure 2 right panel The degree of smoothing is controlled by a smoothing parameter Low values
6. mausmooth Cangemi 2015 and then adding in your reference lists an entry like the following Cangemi Francesco 2015 mausmooth Retrievable online at http phonetik phil fak uni koeln de fcangemi html 4 Where can I find mausmooth Here it is Just copypaste it in your Praat scripting window make all necessary changes run and have fun TITLE MAUSMOOTH Manual and AUtomatic SMOOTHing of f tracks INPUT given wav and potentially Pitch file s ELABORATION plot the f contour in combinations of the three following formats red dots Praat extracted f values Silver dots manually corrected fO points black line smoothed and interpolated f track OUTPUT save plots as single png file then for each sound save manual correction as pitch and smoothed contour as smooth and hesp NOTES The script uses a manual correction pitch file if available otherwise asks user to create a new one pausing the script for each file Working on Praat 5 3 77 under Windows 7 AUTHOR fcangemi uni koeln de INPUT form Input parameters comment path with final slash sentence directory C Documenti temp comment plot parameters sentence figname mausmooth boolean raw 1 boolean manual 1 boolean smooth 1 integer plotmin 75 integer plotmax 400 comment advanced extraction parameters integer smooth1 10 integer smooth2 10 integer timestep integer pitchmin 75 integer pitchmax 600 integer maxcandidates 15
7. mausmootn 1 What is mausmooth mausmooth is a simple script for Praat that helps you extract fO contours from your sound files gives you the Opportunity to make manual changes to the extracted contours and plots them in a smooth way Depending on the aim of your study when plotting fO contours it might be useful to provide information on global trends black line in Figure 1 rather than show the extracted datapoints themselves red dots in Figure 1 which are often affected by various artefacts e g the interactions with segments and phonation types T 4 Figure 1 fO contour for a test utterance plotted using Praat standard extraction method red speckles and mausmooth black line Praat already offers a selection of parameters to minimize the impact of such artefacts But even more reliability might be needed when eyeballing large datasets in order to find differences across categories e g contexts dialects speakers Compare the readability of standard speckle plots left panel in Figure 2 with that of mausmoothed contours right panel in Figure 2 Figure 2 fO contours for utterances from two categories cyan vs magenta plotted using Praat standard extraction and plotting method left panel and using mausmooth right panel 2 How does mausmooth work 2 1 Script basics As the name suggests mausmooth helps you obtain MAnually and AUtomatically SMOOTHed f0 contours It goes through all sound files in a folder
8. readsheet file directory basename hesp ELABORATION 2 PLOT with any form defined combination of original in red corrected in silver smoothed in black Given the order of plotting red surfaces as manually corrected points if raw 1 selectObject Pitch basename Red Speckle plotmin plotmax no endif if manual 1 selectObject Pitch manual Silver Speckle plotmin plotmax no endif if smooth 1 selectObject Pitch smooth Black Draw plotmin plotmax no endif OUTPUT close loops and clean select all minusObject Strings list minusObject Strings pitchlist Remove selectObject Strings list endfor plusObject Strings pitchlist Remove save output Save as 60 dpi PNG file directory figname png
9. tory plot parameters figname mausmooth lv raw lv manual V smooth plotmin plotmax advanced extraction parameters smooth smooth2 timestep pitchmin pitchmax maxcandidates silenceT hr voicing T hr octave octavejump yolceunyoiced Standards Figure 3 mausmooth input form Once the input parameters are validated the script opens the first sound file in the path and checks whether a manual correction a pitch file is already available If this is the case the script plots its fO contour according to the plot parameters If a manual correction is not already available the soundfile is opened in the Sound window the fO contour is extracted and visualized in the Pitch window and the script pauses to let the user provide a manual correction see Figure 4 Pause stop or continue Confirm nn stop Continue A ooo ee ee E 9 Pitch FB_F2_F03 File Edit Query View Selec 0 768073 2888 sa ppe E j B y 7 EET T Md D n ve eur vs a HE Ey 14e i Visible part 1 536145 seconds Total duration 1 536145 seconds gt IY Group Figure 4 Pitch object editor waiting for manual changes At this point the user can use the Selection menu of the Pitch window blue ellipsis in Figure 4 to remove or change spurious pitch points green ellipses in Figure 4 Once changes are made the user can click on Continue in the Pause window red ellipsis in
Download Pdf Manuals
Related Search