Software Requirements Specification
Contents
1. gt Welcome E Terminal Controls gt Training session started gt Device Connect SunSPOT 3B7D Log Data Dump raw data to file gt Waiting for Input x y z out Ye Show Data 01 v Execution Time Terminal In Device Connected Sunspot 3B7D Status Recording Choose Log File lt Reading data from SunSPOT 3B7D This is the window for Training mode c Parameters This panel will allow the user to modify parameters for the training process d Session Controls This panel will allow the user to control the current gesture session i New Session Will create a new gesture session Each gesture session can be saved into a file ii Load Session from File Will load a previously saved session from a file using a standard open file dialog iii Train New Gesture Set Will add a gesture set to the currently opened session Upon clicking on this button Screen 2 2 Train New Gesture will be displayed The ability to train a new gesture set will not be available if the last one created has not yet been saved Appendices Page 21 Software Requirements Specification FROG iv Save Gesture Set Will terminate the collection of instances for a gesture set and then perform the training of the gesture v Exit Training Will terminate the training session close screen 2 0 and open screen 1 0 e Graphical Displays This panel will show different graphical displays such as a graph of 3D acceleration val2. Finish Evaluation Session Will terminate the current evaluation session close i Session Table Will allow the user to select which gestures are to be part of the evaluation as well as display real time statistics ii Sample Size Will allow the user to input the sample size for evaluation iii Randomize If selected will allow the user to be prompted for gestures in a random fashion Otherwise gestures will appear sequentially Appendices Page 26 Software Requirements Specification FROG t Gesture Prompt This panel will display the name and associated image of a gesture as the gesture is requested by the system u Statistics Displays This panel will display recognition evaluation statistics in graphical and tabular format v Connection Panel Refer to Screen 2 0 section j w Terminal Controls Refer to Screen 2 0 section h x Terminal Refer to Screen 2 0 section g Screen 5 0 Demo FROG Demo Game This is the window for Demo mode To utilize Demo mode an appropriate library must be loaded That is Demo mode will require a certain set of predefined gestures to be trained within the library chosen Appendices Page 27 Software Requirements Specification Appendix C Architecture FROG The following represents the FROG architecture It distinguishes system structure as well as data flow Mobile Device Recognition Filtered Vectors Filtering Qu
3. PEN 9 x A 10 5 5 Demo Mods ts id li 10 531 DEM casi til 10 53 2 DEN c 10 5 5 3 DEMOS i 10 A gt DEM ii ic 10 6 JNon functional Requirements nia cine 11 6 1 P rtormanc REUS A o EE pn etr ed uh 11 OLL PRO T 11 RM MEL Li cc RT RS 11 LS PROS 11 6 2 S fety esos n P 11 O SROI Dr EM 11 6 3 Software Quality Requirements ee escribe d bv E FIDA PAR DIM ANTON KIA UN RAM dp DI IONS 11 6 31 QRO M 11 0 3 2 SOR oq o R a 11 6 3 3 IS OUS sanae 11 Appendix A Use Case Mode ia 12 Appendix Br User Interface Prototype iii da 20 Appendix C Arch eot re auo ood ed BUD ERE dum MM MEE 28 Appendix AS ias 29 Table of Contents Page iv Software Requirements Specification FROG 1 Introduction 1 1 Purpose This document gives an overall description of the FROG Project version 1 0 It also describes the requirements both functional and non functional of the FROG Project version 1 0 developed by Team Better Recognize Functional requirements described include those for training recognition evaluation and demonstration Performance safety and software quality are the main non functional requirements in this document 1 2 Intended Audience and Reading Suggestions This specifications document is intended for t
4. 5 2 10 TRA 10 The system shall support real time graphing of 3D accelerometer data from user selected connected mobile devices 52 11 TRA 11 The system shall support logging of system status incoming 3D accelerometer data and execution time of algorithms The user shall be given the capability to save the logged data in a user selected file System Modes Page 8 Software Requirements Specification FROG 5 3 Recognition Mode In this mode the user loads a previously saved training session so as to perform recognition 5 3 1 REC 01 The system shall be able to connect and recognize the gestures of up to four devices simultaneously 5 3 2 REC 02 The system shall allow each user device connected to load a training session library for recognition 5 3 3 REC 03 The system shall provide feedback to each user when a gesture is not recognized or if it is recognized report to the user what gesture it was A non recognition event occurs only when the classifier computes a probability of zero for all gestures in the training session 5 3 4 REC 04 The system shall support logging of system status incoming 3D accelerometer data and execution time of algorithms The user shall be given the capability to save the logged data in a user selected file 5 4 Evaluation Mode In this mode the user may again perform recognition but the system shall also provide useful information for evaluating performance 5 4 4 EVA 01 The sy
5. Software Requirements Specification FROG Appendix A Use Case Model Recognize Pd User Connect Main Success Scenario User selects Configuration option System detects and displays available devices with a unique meaningful ID User can select one of the available devices to connect to System formally connects to the device and adds it to a list of connected devices User can select filters or use the default filters User can disconnect and repeat from step 3 User selects to leave Configuration option APA qa tB pa Appendices Page 12 Software Requirements Specification FROG Demo Mode Precondition Gestures to be recognized have been previously trained Main Success Scenario User selects Demo mode System displays demo configuration window User s perform connection and select which libraries to use User selects to begin game System displays demo window and starts the game User s follow the on screen instructions and must make use of gestures to win the game 7 System displays game over 8 User selects to leave Demo mode or play again e Extensions 4a User selected a gesture session that did not contain the demo s required gestures 1 System halts and informs user exactly what gestures are needed Appendices Page 13 Software Requirements Specification FROG Evaluation Mode Precondition Gestures to be recognized have been previously trained Main Success S
6. Screen 4 0 Evaluation Mode FROG Recognizer Of Gestures Evaluation Session Settings Gesture Library Loaded BasicShapes ag Load New Library Finish Evaluation Session Customize Evaluation Session Sample Size 15 Randomize Gesture Correct Incorrect Av Certainty Circle Square 50 O Star Triangle Correct Incorrect Unrecognized Live Statistics Show in new window gt Welcome gt Setting Evaluation session gt Device Connect SunSPOT 3B7D gt ID Josh gt Loading Gesture library 100 gt Library Loaded BasicShapes ag gt Starting Evaluation session gt Gesture Recognized triangle gt right gt Gesture not recognized sorry gt Gesture Recognized circle gt wrong sorry lt iil Recognition Ignore Last gesture D Device Set Up Device Sun Spot Threshold 0 25 Terminal Controls Log Data Communication Data Choose Log File Show Certainity of Rec Excecution Time New Window Device Connected SunSPOT 3B7D Status Recording Reading Data from Sun Spot 3B7D This is the window for Evaluation mode r Session Settings This panel will allow the user to load a library or finish the session i Load New Library Refer to Screen 3 0 section m Screen 3 0 and return to Screen 1 0 s Customize Evaluation Session This panel will allow the user to modify the parameters of the evaluation session
7. k means algorithm and quantizes the vectors 4 System uses the Bayesian classifier to compute the probability of the incoming gesture matching previously trained gestures in the library 5 System returns that the gesture was recognized as the gesture with the highest classified probability or unrecognized if no match has probability greater than zero Recognize Appendices Page 16 Software Requirements Specification FROG Recognition Mode Precondition Gestures to be recognized have been previously trained Main Success Scenario User selects Recognize mode User s performs connection User s select s to load a new library User s make the gesture previously trained while holding the send button on their device System responds with the name and the image of the gesture it recognized User can display system messages acceleration data and execution time in the terminal or log it into a text file 7 User s can repeat steps from 2 3 or 4 8 User selects to leave Recognition mode ae eee E Extensions 6a System did not recognize the gesture 1 System responds to user with an appropriate unrecognized response Appendices Page 17 Software Requirements Specification FROG Main Success Scenario 1 System accepts raw acceleration data from the user 2 System applies user chosen filters on the data 3 System applies k means algorithm on the current and previously filtered acceleration data to obtain quant
8. Then formally stated the Baum Welch algorithm is Algorithm Baum Welch Input A set of observed sequences O 0 Initialization Select arbitrary model parameters N 0 5 i score Y P O A Repeat 1 A A S5 S For each sequence Of Calculate probable paths Q qd qd Calculate a t i for O using the Forward algorithm Calculate t i for O using the Backward algorithm Calculate the contribution of O to A using 1 Calculate the contribution of O to E using 2 Aij ai SN As Ej eie Eh score 4 P Olai es j Until the change in score is less than some predefined threshold Where ta l 1 Ay Y POF X a t i aijzes Of 1 A t 1 i d t 1 Appendices Page 30 Software Requirements Specification FROG Blo 3 ko M a ti 8 t i t 1 Od o O is the sequence of d observations of the complete observation sequence a and f refer just as they did above to the respective forward backward algorithm partial probabilities And so with each iteration the HMM model parameters are modified towards a local optimum training the model to the sequences provided Appendices Page 31
9. algorithm is defined recursively as Br t 1 Be1 G Xi AB a 5 E bli oz As it is exactly derivative of the forward algorithm above only now using the p function further explanation will not be given here Appendices Page 29 Software Requirements Specification FROG Baum Welch Algorithm The Baum Welch algorithm deals instead with the training problem The Baum Welch is an expectation maximization algorithm that utilizes the forward and backward algorithms The algorithm is designed to optimize the parameters of an HMM so as to best model given training sequences Thus it deals with maximizing the conditional probability of an observation sequence occurring given an HMM to be optimized The algorithm is only capable of carrying out local optimization however so there is no guarantee of the truly optimal HMM which would require knowledge of a global optimum The algorithm we implement differs slightly from the generalized definition of Baum Welch because we process our data as a left right model One has to use a multiple observation sequence method as prescribed by Rabiner 273 The modified method is required due to the rigid nature of a left right model which can lead to dramatic overtraining if carried out on an individual instance basis This modified algorithm is detailed below Here let a again be the transition matrix but use notation e 0x to represent the emission probability for observation k from state 1
10. antizer Quantize K means Translated Vectors Match Probability Classifier Classify Recognition Data Session Object Y Filter Parameters Y 41d Connection Communication Gesture Session User Interface Log File Plug in ba 9 File Library r A A E s e 2 e 2 5 S S S a g p a o E o o 8 E 5 E 8 8 S S 3 8 5 A E 2 s SIB z E E E J 8 g i 6 z i Hi S 3 N Training K means Translated Vectors Quantizer Model Filtering Filtered Vectors Appendices Page 28 Software Requirements Specification FROG Appendix D Algorithms This section describes algorithms necessary for gesture recognition in the FROG project Forward Algorithm The forward algorithm is designed for computing the probability of observing a certain sequence of observations being emitted from an HMM In other words it deals with the problem of recognition Its viability comes from its marked increase in efficiency as compared to that of a brute force approach A brute force approach would involve traversing every possible state path and combining each path s probability of producing the sequence The forward algorithm on the other hand utilizes the Markov property of an HMM to drastically improve this
11. cenario l 2 3 4 DU 9 10 11 12 User selects Evaluation mode User performs connection User can select to load a new library User selects evaluation parameters sample size and random or sequential gesture prompting System displays a gesture User makes that gesture System displays information about the instance performed and keeps a running total of attempts to make a gesture User can export data collected on evaluated gestures such as the number of correct and incorrect gestures as well as the average certainty User can also view graphical information live about the above mentioned statistics User can make the system display system messages acceleration data and execution time in the terminal or log it to a text file User can repeat from steps 2 3 or 5 User selects to leave Evaluation mode Appendices Page 14 Software Requirements Specification FROG Load Library Main Success Scenario 1 User chooses to load a library 2 System displays a file system window that gives the user the option of opening any previously saved library 3 System loads the gesture images gestures names and training data from the library Appendices Page 15 Software Requirements Specification FROG Main Success Scenario 1 System accepts raw acceleration data from the user 2 System applies user chosen filters on the data 3 System compares the filtered acceleration data to computed centers for the
12. h both a User Manual and Developer Manual These documents and more will be delivered to the project sponsor on a DVD on completion of the project The User Manual will be a step by step guide that can walk a user through the installation and operation of the system The Developer Manual will help those who might be developing a plug in or making modifications to the existing system 3 7 Assumptions and Dependencies The FROG Project assumes the following e The end user has a Java Virtual Machine compatible with their platform e The end user has a background in gesture recognition or is at least somewhat familiar with gesture recognition technology so as to facilitate proper use of the product Overall Description Page 5 Software Requirements Specification FROG 4 External Interface Requirements 4 1 User Interfaces The user interface shall be clean and intuitively labeled to promote a high quality look that users will find easy to use Each window shall contain an output window pane that will show all status messages and data being processed The user shall have the ability to select desired console and display output There shall also be a connection panel in each mode that allows the user to see the status of their device s connection with the system as well as help them connect reconnect disconnect the device 4 2 Software Interfaces The project shall have a plug in based system for adding support for additional mobile device
13. he customer and developers of this project It ensures that the customer and Team Better Recognize are better synchronized and share the same vision 1 3 Product Scope The FROG Project has as its goal the development of a device and platform independent gesture training and recognition system In its initial release FROG will be packaged with a plug in for Sun SPOTS Additional plug ins may be written later based on a plug in framework In addition FROG will contain an evaluation mode as well as a demo program for testing and demonstrating the capabilities of the project 1 4 References Hobermann Rose Durand Dannie HMM Lecture Notes 2006 Carnegie Mellon School of Computer Science 10 September 2009 http www cs cmu edu durand 03 711 2006 Lectures hmm bw pdf Rabiner L R A tutorial on hidden Markov models and selected applications in speech recognition Proceedings of the IEEE 77 Feb 1989 257 286 Schl mer Thomas Poppinga Benjamin Henze Niels Boll Susanne Gesture Recognition with a Wii Controller 2008 http wiigee org 10 September 2009 Introduction Page 1 Software Requirements Specification FROG 2 Definition of Terms Accelerometer Gesture Hidden Markov Model K means Clustering K means Clustering Sun SPOT An instrument for measuring acceleration In particular a 3D accelerometer measures acceleration in three dimensions A continuous combination of motions made by an individua
14. ition Evaluation and Demo The Training mode will allow the user to train gestures for later use The Recognition mode will allow the user to load previously trained gestures and perform gesture recognition using that library The Evaluation mode allows the user to determine recognition accuracy as well as view other performance statistics to diagnose issues with their hardware or the FROG system itself The Demo mode is a simple game to demonstrate the capabilities of the system 3 3 User Classes and Characteristics This product is being developed for use in an experimental academic environment This product is designed for use by anyone who understands how to operate a PC or Mac and reads the FROG User Manual The demo is meant to be a fun and light hearted way for anyone to use the system 3 4 Operating Environment The FROG Project was designed to operate with the following software installed e Windows XP or later Mac OSX or Linux e Java Runtime Environment 6 0 or later 3 5 Design and Implementation Constraints Time Constraint Limited by academic school year ending on May 11 2010 Mobile Device Limitations J2ME library on Sun SPOTs Data communication between mobile device and host may be limited Hardware Limitations System developed on older machines this may cause performance issues Overall Description Page 4 Software Requirements Specification FROG 3 6 User Documentation The complete FROG Project will come wit
15. ized vectors 4 System applies the Baum Welch algorithm to create an optimized HMM Train Appendices Page 18 Software Requirements Specification FROG Training Mode Main Success Scenario l 2 B 10 11 12 13 14 15 16 User selects Training mode User performs connection User chooses to train a new gesture set which will require the naming of the gesture as well as the optional association of an image with the gesture User can choose to either keep the default training settings or change training parameters such as sampling frequency as well as number of centers for k means and states for the HMM User can choose a file to dump raw or filtered acceleration data for further use User makes a gesture while holding the send button on their device User can display system messages acceleration data and execution time in the terminal or log it into a text file User can view the different graphical displays that plot graph vectors User may choose to edit the gesture allowing them to view and delete individual instances as well as change the associated image While want more instances repeat from step 6 If there are no more instances the user chooses to train the gesture set User can delete a gesture While more gestures repeat from step 3 User selects to leave Training mode or create a new session System prompts the user to save session System prompts for name location of save file for session A
16. l technology by Sun Microsystems They contain a 180MHz 32 bit processor 512K RAM and a variety of sensors including a three axis accelerometer used in the FROG project Sun SPOTs communicate using a low power IEEE 802 15 4 radio Definition of Terms Page 2 Software Requirements Specification FROG Training Instance A training instance is a single motion of the mobile device by a user representing a gesture to be trained Training Session A training session is a collection of training sets created by a user A session is saved in a file format and reloaded to perform recognition A training session can be thought of as a project file containing representations of multiple gestures Training Set A training set is a sequence of training instances created by the user in order to train a gesture A training set can be thought of as the complete set of data used to create an HMM representation of a gesture Definition of Terms Page 3 Software Requirements Specification FROG 3 Overall Description 3 1 Product Perspective FROG is a self contained gesture recognition system It is to be used in conjunction with plug ins that contain device specific code so that it has potential compatibility with any wireless 3D accelerometer enabled mobile device 3 2 Product Functions The FROG Project will be a 3D acceleration based gesture training and recognition system consisting of four main modes These will be Training Recogn
17. l usually with the hands that are related and meaningful as a whole Gestures are the entities which shall be modeled as well as recognized by the FROG project A doubly stochastic as opposed to deterministic math model being used to represent a gesture Constructed HMMs are then used in recognition A statistical model in which the system being modeled is assumed to be a Markov process with unobserved state This document uses the convention that an HMM is defined as S O a b x where S The set of hidden states O The set of possible observations here dealing with vectors a The transition matrix where a i j represents the probability of transitioning from state i to state j b The emission probability matrix where b i k represents the probability of emitting observation k while in state 1 n The initial condition probability distribution for initial state Method of cluster analysis which aims to partition n observations into k clusters in which each observation is clustered with the nearest mean Method of cluster analysis that carries out the exact same algorithm for clustering as k means However k means chooses initial conditions based on the input instead of arbitrarily as in k means K means is designed to be more efficient and accurate than k means Sun SPOTS Sun Small Programmable Object Technology are small programmable wireless sensor devices developed as an experimenta
18. osh gt gt Square 50 certain Jose gt gt Circle 20 Device SunSPOT2C06 w y Threshold 0 25 v Load New Library Device Connected SunSPOT 2C06 Status Recording Slot 04 Q Device Select RA Use the Device menu at Threshold x i Mee EB The right to set a device 10 25 Y To work with this slot Load New Library X No Device Connected Status Device Not Found Exit Recognition Q Terminal Controls v Log Data 9 a Choose Log File Communication Data Show Certainity of Rec Terminal In New Window Recognition session started This is the window for Recognition mode m User Panel This panel will handle one user There will be four of these slots in Screen 3 0 Each one will allow the following actions i Connection Panel Refer to Screen 2 0 section j ii Load New Library Will allow each user to select his or her own library gesture session file to be used for recognition Upon clicking this button the user will be prompted with a standard open file dialog n User Panel Unconnected This User Panel shows the appearance of such a panel when no user is connected o Exit Recognition Will terminate the current recognition session close Screen 3 0 and return to Screen 1 0 p Terminal Controls Refer to Screen 2 0 section h q Terminal Refer to Screen 2 0 section g Appendices Page 25 Software Requirements Specification FROG
19. ppendices Page 19 Software Requirements Specification FROG Appendix B User Interface Prototype The following is a prototype of the FROG user interface This prototype gives a screenshot based walkthrough of the modes and use of the FROG Recognizer of Gestures product Screen 1 0 Main Menu FROG Recognizer Of Gestures Welcome to FROG FROG Recognizer Of Gestures Train Recognize Evaluate Demo o This is the main window of the program From this window the user can access the different modes In order for this window to display it is necessary for the system to find at least one device plug in to work with If no plug in is encountered then the program will ask the user to install a plug in a This button will display information about the program when clicked b Exit button Appendices Page 20 Software Requirements Specification FROG Screen 2 0 Training Mode FROG Recognizer Of Gestures Training Parameters Session Controls Load Session From File Train New Gesture Set Save Gesture Set as Device Threshold Filter s Advanced SUNSPOT2C06 0 25 Sampling Freq select Filter Y Advanced Quantizer K Value Select Filter Advanced Hmm States select Restore Defaults Vectors Display Other Graphical Display dl Exit Training 1 Triangle o y X 1 2563 Y 5 2365 Z 2 8569 Y
20. process The algorithm relies on calculating partial probabilities at each time step in the observation sequence that is consider each observation in the sequence as arriving at a discrete time t corresponding to the place it occurs in the sequence This document defines the partial probabilities recursively using the notation defined in Section 2 A S O a b v 2 j TM Et 0 a i j oq Ti cbli The variable o represents the observation occurring at time t And so the algorithm is based on initializing the a i according to the initial probability multiplied by the emission probability of the first symbol in the observation sequence Then o41 j is the emission probability for the next observation multiplied by the sum of all the previous partial probabilities multiplied by the probability of transitioning from each state to the new state j The recursive computation of this value can yield the probability of observing any sequence by simply summing up over the ar i where T is the total number of observations in the sequence that is sum over the last set of partial probabilities computed Backward Algorithm The backward algorithm is based on exactly the same premises as the forward algorithm It is also utilized for calculating the probability of an observed sequence from an HMM This algorithm however calculates probabilities as suggested by its name starting from the last observation and working backward The
21. s The plug ins shall contain all device specific code and any translation or handling needed for accelerometer data collected from the device The plug ins shall be derived from a common interface within the FROG software Since it is written in Java FROG will have an obvious need to interact with a Java Virtual Machine Standard Edition 6 0 or later External Interface Requirements Page 6 Software Requirements Specification FROG 5 Functional Requirements 5 1 Requirements for All Modes 5 1 1 GEN 01 The system shall take 3D accelerometer readings from mobile devices as its input From this data the system shall perform its gesture training and recognition 5 1 2 GEN 02 The system shall not include internal device specific support Support for each mobile device shall be incorporated into a corresponding plug in for the device 5 1 3 GEN 03 The system shall provide a console window or pane at all times to display relevant user selectable information 5 2 Training Mode In this mode the user records gestures names them and saves them to be used later in other modes The user creates a series of training instances for each training set The user s training session will then be composed of the collection of these training sets 5 2 1 TRA 01 The system shall limit the number of connected devices to one during training mode 5 2 2 TRA 02 The system shall provide the user with the ability to save and load training ses
22. sions for reuse The file the system creates will be platform independent 5 2 3 TRA 03 The system shall provide a user with an intuitive method of training gestures into the system These gestures may be represented by a word and a picture or illustration System Modes Page 7 Software Requirements Specification FROG The system shall support a display of available trained gestures in a particular library 5 2 4 TRA 04 5 2 5 TRA 05 The system shall allow the user to load an existing training session either to add additional gestures training sets to the file or to delete gestures training sets permanently from the file 52 6 TRA 06 The system shall support a filtering framework with idle state and directorial equivalence filters as defaults based on the Wiigee project The framework will support the addition and modification of filters for each training instance 5 2 7 TRA 07 The system shall support gesture training through vector quantization and HMM training Both k means and k means will be supported for vector quantization 52 8 TRA 08 The system shall support modification albeit limited to certain values of the number of centers k in the k means k means algorithm with a default of k 14 based on the Wiigee project for each training set 5 2 9 TRA 09 The system shall support modification of the number of HMM states with a default of 8 states based on the Wiigee project for each training set
23. software Requirements specification FROG Recognizer of Gestures Team Better Recognize Version 3 0 April 30 2010 FROG Copyright 2009 2010 Computer Science Department Texas Christian University Software Requirements Specification FROG Revision Sign off By signing the following the team member asserts that he she has read the entire document and has to the best of his or her knowledge found the information contained herein to be accurate relevant and free of typographical error Name Signature Date Josh Alvord Alex Grosso Jose Marquez Sneha Popley Phillip Stromberg Ford Wesner Revision Sign off Page i Software Requirements Specification FROG Revision History The following is a history of revisions of this document Document Version Date Edited Changes Version 1 0 11 03 09 Initial Draft Version 1 1 11 19 09 Specified corrections in content Version 2 0 2 2 10 Iteration 2 Update Version 3 0 04 30 10 Final Iteration Update Revision History Page ii Software Requirements Specification FROG Table of Contents Revision Sien x aetema muda edad eden uite aa ded e i Revision History cs o Doa ea RS dico M So a HN EM MS NE 11 EM A Tc 1 1 1 PUSS T M 1 1 2 Intended Audience and Reading Suggestions eese 1 1 3 Product SCOPE dos 1 E NES A o
24. stem shall allow only one device to connect at a time 5 4 2 EVA 02 The system shall provide real time feedback of the performance of the recognition system This shall include a tally of number correct number not recognized and number matched incorrectly as well as average certainty 5 4 3 EVA 03 The system shall provide the ability for the user to input the sample size number to be requested for recognition for each gesture from the library Each gesture may be enabled or disabled individually and they will be prompted either sequentially or at random based on the user s choice System Modes Page 9 Software Requirements Specification FROG The system shall support logging of system status incoming 3D accelerometer data and execution time of algorithms The user shall be given the capability to save the logged data in a user selected file 5 4 4 EVA 04 5 5 Demo Mode 5 5 1 DEM 01 The system shall allow up to four devices to connect and play 5 5 2 DEM 02 The system shall contain at least one client approved demo program to better showcase the abilities of the underlying recognition system 5 5 3 DEM 03 The system shall keep track of each user s score correct gestures made for evaluation of the system s or perhaps the users performance 5 5 4 DEM 04 The system shall require each user to have trained the gestures needed for the demo The users currently loaded training session file must contain gestures tha
25. t share the same names as defined by the demo A message to this effect must be displayed to the user so that he she may go back to training mode and correct the situation System Modes Page 10 Software Requirements Specification FROG 6 Non functional Requirements 6 1 Performance Requirements 6 1 1 PR 01 Communication between host and device shall be fast enough to support up to four connected devices Recognition performance must not slow by more than 20 with four users connected 6 1 2 PR 02 The speed at which a gesture is recognized shall not exceed 10ms per traditionally sized HMM 8 hidden states 14 observable states 6 1 3 PR 03 Code written for the Sun SPOT i e for filtering shall not hinder its ability to process accelerometer and radio transmission reception data Hinder is defined here as causing the threads associated with the above activities to be skipped for more than 1 period 6 2 Safety Requirements 6 2 1 SR 01 Accelerometer enabled mobile devices must be used with attention to surroundings The user must not allow the device to become airborne potentially causing injury or damage 6 3 Software Quality Requirements 6 3 1 SQR 01 The FROG Project shall be able to recognize gestures with an 80 or better accuracy 6 3 2 SQR 02 Plug ins shall be usable to extend the versatility of the software 6 3 3 SOR 03 FROG shall be a multi platform framework Non functional Requirements Page 11
26. the text file ii Terminal in New Window Will display the terminal in a separate window for easier manipulation see Screen 2 1 i Device Status Display This panel will let the user know the current status of the mobile device used to input data to the system j Connection Panel This panel will give the user the option of setting up a device to work with the system This panel will appear with a different configuration in other windows and will allow the user the following capabilities i Device Will allow the user to choose the kind of device to be set up The choices of this combo box will correspond to the plug ins installed ii Threshold Will allow the user to prescribe a threshold to be used with recognition performed with that mobile device ii Connect Will utilize the device s corresponding plug in to carry out connection of the mobile device If any further user action is necessary the plug in is responsible for requesting such actions from the user Appendices Page 22 Software Requirements Specification FROG Screen 2 1 External Terminal FROG Console gt Welcome Log Data gt Training session started Device Connect SunSPOT 3B7D x y z out Waiting for Input Show Data 01 Execution Time Dump raw data to file Choose Log File This window is an external terminal It will display the output given to the user by the mode currently running k Terminal Refer
27. to Screen 2 0 section g D Terminal Controls Refer to Screen 2 0 section h Appendices Page 23 Software Requirements Specification FROG Screen 2 2 Train New Gesture FROG Train New Gesture This window will allow the user to name the new gesture to be created and will also give the user the ability to attach a symbolic image to the gesture for easier identification of the gesture s semantic meaning in other modes Appendices Page 24 Software Requirements Specification FROG Screen 3 0 Recognition Mode FROG Recognizer Of Gestures Recognition ID Josh ID Ford Library RoundShapes ag Library BasicShapes ag Device SunSPOT3B7D w X Threshold 0 25 x Load New Library Device Connected SunSPOT 3B7D Status Recording ID Jose Library ComplexShapes ag Device SunSPOT 415E w Threshold 0 25 v Load New Library Device Connected SunSPOT 415E Status Recording gt Welcome gt Setting Recognition session gt Slot 01 Device Connect SunSPOT 3B7D gt Slot 02 Device Connect SunSPOT 2C06 gt Slot 03 Device Connect SunSPOT 415E gt Slot 01 Loading Library BasicShapes ag Session Started for Josh gt Slot 02 Loading Library RoundShapes ag Session Started for Ford gt Slot 03 Loading Library ComplexShapes ag Session Started for Jose gt Starting Recognition session Ford gt gt Circle 99 certain Jose gt gt Gesture not recognized sorry J
28. u ee es 1 EN D imt on of Terms A AP aA aaa i an 2 De Overall Descriptio erene oL DRE 4 3 1 Product Persp ct iaa 4 3 2 A NT 4 3 3 User Classes and Characidae 4 3 4 Operating BUVIFODISEHL s uides edad at UA isrener EROR E ROSE ED UMEN RR DAP FU DIN REESS HM E IDE Ud 4 3 5 Design and Implementation Constraints A 4 3 6 User TS SG To LE 5 3 7 Assumptions and Dependencies iecore ee ber tied chop on ER REED ERE dt 5 4 Exte rnal Interface Requires sii 6 4 1 User IMETA CES A RAE E E RE AEE RE RE EEOAE 6 4 2 Software Interfaces testcase utes orner dd cae eE a N 6 5 Functional REQUIEM 7 5 1 Requirements tor All Mods ii ean 7 A o o 7 LL DEN iS T la GEN O3 ora T D drang Mode att 7 SAL TRA Oleee tapis eee A 7 o PRAO taa 7 32 3 TRAO its 7 52A ERA O iS 8 5 2 3 TRAJ TT M 8 dabo MU M ME DM A KM UM 8 WMV I id 8 529 TERA eo ene te arene ream Ns 8 Table of Contents Page iii Software Requirements Specification FROG NEU T LO C 8 SUNL MER o 1d fr ere 8 SLE ZEND esos ce Dd HD E 8 5 3 BR ECO STIG OI MO M 9 PLUME aci c PPP E 9 93 2 REC o ae 9 a EE i e io 9 334 RECO c E 9 s ME lucido o m 9 TAL EA 9 VE WM D c aa 9 M OMERI S Ci
29. ues as well as the clusters from the k means algorithm f Session Panel This panel will show the gestures that are part of the current gesture session and allow the user to load an image using a standard open file dialog to represent each gesture by clicking on the icon to the right of the current image g Terminal The terminal will serve as the primary method of providing feedback to the user in real time about the actions being performed performance etc To the side of the terminal there will be controls for the user to select the feedback needed see h This description applies to the terminals found in other modes h Terminal Controls The terminal controls will allow the user to choose the feedback he or she wants to appear in the terminal In Training mode there will be the option of displaying the acceleration values of each axis and performance of the algorithms involved More controls may be added during the design process Some other actions that those controls allow the user are i Dump Raw Data to File Will allow the user to choose a text file where the acceleration data received from a connected mobile device will be dumped Note that this capability is exclusive to the Training mode screen ii Choose Log File Will allow the user to save the contents being posted in the terminal to a text file Upon clicking this button the user will be prompted with a standard save file dialog for the user to select the location and name of
Download Pdf Manuals
Related Search