Tactile Hearing Aid Demonstration ENSC 440 – Group 14 David Dickin Mehran Eghtesad Ryan Dickie Merle Kinkade Simon Fraser University April 2007
Copyright 2007 Pivit Technologies Inc.2 Outline Introduction Software Subsystems Hardware Subsystems Business Case Experimental Results Conclusion Demonstration
Introduction
April 2007Copyright 2007 Pivit Technologies Inc.4 Team Members David Dickin – CEO Mehran Eghtesad – Senior Hardware Designer Ryan Dickie – Senior Software Designer Merle Kinkade – CFO
April 2007Copyright 2007 Pivit Technologies Inc.5 What is a Tactile Hearing Aid? A device that translates noise into touch Used to help the hearing impaired Traditional hearing aids cannot be used by all people
April 2007Copyright 2007 Pivit Technologies Inc.6 System Overview
April 2007Copyright 2007 Pivit Technologies Inc.7 Subsystems
Software Subsystems
April 2007Copyright 2007 Pivit Technologies Inc.9 Technologies Wav Reader/Writer: self-designed. FFT: FFTW (Fastest Fourier Transform in the West) Audio Input/Outer: Windows Multimedia Library Parallel Port: Inpout32.dll Drawing: GDI GUI : Winapi Misc: Boost C++ libraries
April 2007Copyright 2007 Pivit Technologies Inc.10 GUI User Interface
April 2007Copyright 2007 Pivit Technologies Inc.11 Audio IO Runs in separate threads Hz, 16-bit, mono Can load/save to wav files Can use microphone input
April 2007Copyright 2007 Pivit Technologies Inc.12 DSP Hamming Windows R2C DFT with N = 4096 Threshold Filtering Critical Transform Bins Buzzer Amplitude Mapping
April 2007Copyright 2007 Pivit Technologies Inc.13 Parallel Port Interface Runs in a separate thread Updates in 1ms intervals Buzzer amplitude achieved by pulsing buzzers Uses inpout32.dll driver
Hardware Subsystems
April 2007Copyright 2007 Pivit Technologies Inc.15 I/O Control Board PCB fabricated to drive vibrators Connects software to VibraPad UI is on/off switch and intensity knob
April 2007Copyright 2007 Pivit Technologies Inc.16 VibraPad Why 16 Vibrators? Parallel port limitations Space limitations Power restrictions Worn on stomach/torso area Largest even surface area without contour issues Large network of nerve endings Velcro for repositioning and ease of array placement
April 2007Copyright 2007 Pivit Technologies Inc.17 Proposed Mappings Uniform Offset Offset with Snaking Offset with Snaking & Natural Mapping
April 2007Copyright 2007 Pivit Technologies Inc.18 Transforms/Mapping Critical Bands based on Frequency Response of Human Ear Critical Band Mapping based on Bark Scale Below 500Hz Uniform bands of 100Hz intervals Above 500Hz Bands increase non- uniformly with Frequency
April 2007Copyright 2007 Pivit Technologies Inc.19 Bark Scale Divisions High resolution below 500Hz Human voice ~500Hz–2kHz Optimum range ~2kHz– 4kHz Resolution decreases Rapidly above 4kHz
Business Case
April 2007Copyright 2007 Pivit Technologies Inc.21 Budget and Funding
April 2007Copyright 2007 Pivit Technologies Inc.22 Schedule
April 2007Copyright 2007 Pivit Technologies Inc.23 Market Overview Statistics obtained from publications of the National Institute on Deafness and other Communication Disorders (NIDCD) based on the American population and published at
Experimental Results
April 2007Copyright 2007 Pivit Technologies Inc.25 Goals “The purpose of this project is to determine the extent and limitations of the tactile sensory system to receive and translate sound information. Such measures will be quantified by outlining specific test procedures and conducting these tests on several individuals.” Project Proposal Criteria for Success “To deem a series of experiments successful, a minimum hit rate of 80% must be achieved after a maximum of 100 minutes of cumulative training, or 5 iterations of the above experiment.” Design Specifications
April 2007Copyright 2007 Pivit Technologies Inc.26 Experiment 1 Procedure 10 short sound clips Sirens, bells, animals, vehicles, etc. Subjects allowed to hear sounds Training time One by one with audio: open ended One by one without audio: open ended Random: max 10 minutes Operator-assisted training: max 10 minutes Two set of Transforms
April 2007Copyright 2007 Pivit Technologies Inc.27 Transforms Transform 1: Frequency bins determined based on critical bands Transform 2: Frequency bins divided equally in log scale BuzzerFrequency Range (Hz) Transform 1Transform
April 2007Copyright 2007 Pivit Technologies Inc.28 Experiment 1 Results Conducted on 4 people Transform 1: Average Hit rate: 95% Average Training time: 20 min PASS Transform 2: Average Hit rate: 100% Average Training time: 16 min PASS TOO EASY?
April 2007Copyright 2007 Pivit Technologies Inc.29 Experiment 2 Procedure Changes from Experiment 1 30 short sound clips Normalize length of clips (3-4 sec) Subject not allowed to hear sounds
April 2007Copyright 2007 Pivit Technologies Inc.30 Experiment 2 Results Conducted on 4 people Transform 1: Average Hit rate: 55% Average Training time: 43 min FAIL Transform 2: Average Hit rate: 53% Average Training time: 35 min FAIL TOO HARD?
April 2007Copyright 2007 Pivit Technologies Inc.31 Redesign Changes Based on user feedback and comments as well as our personal observations Create Transform 3: compromise of uniform and critical band Change VibraPad layout Change different intensity levels (PWM)
April 2007Copyright 2007 Pivit Technologies Inc.32 Vibrator Layout Experiment 1/2 LayoutExperiment 3 Layout
April 2007Copyright 2007 Pivit Technologies Inc.33 Experiment 3 Procedure Large scale test (20+ subjects) 10 short sound clips of the same length Sirens, bells, alarms, cars, etc. Limited training time Self directed: 7 minutes Operator-assisted: 7 minutes Provide a list of sound names to subjects Using transform 3
April 2007Copyright 2007 Pivit Technologies Inc.34 Experiment 3 Results Conducted on 23 individuals Average hit rate: 68% Average training time: 17.5 min
April 2007Copyright 2007 Pivit Technologies Inc.35 Experiment 3 Results Previous exposure Our group: 97.5% Everyone else: 61% Age 16 Individuals under 35: 77% 7 individuals over 35: 44%
Conclusion
April 2007Copyright 2007 Pivit Technologies Inc.37 Summary Participants are able to identify sounds with practice High pitched sounds, like alarms, are easy to distinguish More testing required, but has some marketability Project stayed on schedule and budget
April 2007Copyright 2007 Pivit Technologies Inc.38 Acknowledgements Thanks to the ESSEF and the Wighton Fund for funding for our project Thanks to the ENSC 440/305 Professors and TA’s for their time and assistance
April 2007Copyright 2007 Pivit Technologies Inc.39 Questions? Go Canucks!
April 2007Copyright 2007 Pivit Technologies Inc.40 Similar Devices Audiotact Tacticon Tactaid
April 2007Copyright 2007 Pivit Technologies Inc.41 Bark Scale Divisions
April 2007Copyright 2007 Pivit Technologies Inc.42 Standards Standards Council of Canada CSA ANSI ISO FCC ECMA