Post on 31-Dec-2015
description
Anti-Snoring Pillow (ASP)Anti-Snoring Pillow (ASP)
December 13, 2007December 13, 2007
For a peaceful night of sleep
LifeX Team
Raymond Lee Software Researching parts
Camillia Lee Documentation Software Testing
Simon Wong Theory Software Debugger
Stanley Yang Software Budget
OutlineBackgroundObjectivesSystem OverviewHigh Level System DesignBusiness CaseResultsWhat was learnedFuture ImprovementsConclusion
Background
Background
“Forty-five percent of normal adults snore at least occasionally, and 25 percent are
habitual snorers.”
“Thirty percent of adults over age 30 are snorers. By middle age, that number
reaches 40 percent.”
Background… continued
A number of effects to both the snorer and those who hear him/her
daytime drowsiness, irritability, lack of focus, decrease libidopsychological and social damage
Current existing solutions
Surgeries, sleeping aids, dental appliances
Downfalls Expensive Invasive Painful Complications Unreliable
Objectives
Objectives
Produce a affordable non-invasive solution to reduce the sound of snoring
Goal: Minimize snoring noise at low frequencies by 10-15dB
LifeX’s Solution
The “Anti-Snoring Pillow”
-A noise suppression system integrated into a pillow
System Overview
Types of Noise Control - Passive
Reduces noise using specialized materials Sound isolation Sound absorption Vibration damping
i.e. Ear muffs
Types of Noise Control - Active
Acoustic cancellation that involves a control speaker for emitting a opposite polarity sound
Adaptive ANC
Adaptive ANC Real time controller for monitoring the system’s
performance System parameters are always changing Required for complex noise (i.e. speech,
snoring, random noise, etc)
Adaptive ANC
How? Digital signal travels faster than speed of
sound!Advantages over passive acoustic control
More effective at low frequencies Less bulky Able to block noise selectively A “good” system will yield better performance
(up to 20+dB reduction) Adaptive!!!
System Overview
1x Speaker (Control)2x Microphone (Reference & Error)1x DSP board1x Pillow
System Arrangement
High Level System Design
Active Noise Cancellation Systems
Types of ANC system Digital Filters Adaptation Algorithm
Types of ANC System
Two Major types Waveform synthesis (Periodic noise – Engine
noise, fan noise) Adaptive Filtering
Feedback (No reference signal) Feedforward (Reference signal)
• Feedforward is always preferred over feedback when reference signal is available
High Level System Design
Feedforward System
Adaptive broadband feedforward control with an acoustic input sensor
Digital Filters
Finite Impulse Response (FIR) Inherently stable
Infinite Impulse Response (IIR) Built in feedback compensation Less computational low Can model complex systems
Inherently unstable
Digital Filters
Three major parameters: type of system, filter weights, number of filter weights Optimization by trial and error
Adaptation Algorithm
Least Mean Square (LMS) FXLMS
Secondary path compensation (Offline Training)
Adaptation Algorithm
Filtered-U Recursive (RLMS)
Business Case
Market Our target market would be towards couples sleeping on
the same bed
Our anti-snoring product is unique compared to other solutions available
Benefits to our product Non-invasive Inexpensive Safe Comfortable User friendly
CostParts (in thousands)
TI DSK 6713 $20,000
Microphones x 2 $7,000
Speakers x 2 $60,000
Pillow $30,000
Analog parts $1,000
Parts Total $132,000
Services
Packaging $1,000
Labour $9,000
Market Fees $1,000
Market agent's fees $3,000
Service Total $14,000
Total Cost $146,000
Total Revenue (1000 x $200) $200,000
Total Profit $78,000
Financing
Bank loans Investment banking
Private investorsAngel investors
Competition
High performance passive ANC foam ear plugs
Chin-up Strips Keeps mouth closed to reduce snoring
Nasal strips Keep nostrils opened for better breathing
SurgeryNone using Active Noise Cancellation!!!
Results
Snoring Sample Spectrum
Experimental Results – 1st Try
Simplified approach…
Results
Sine waves
Frequency (Hz) Attenuation (dB)
200 ~ 10 dB
300 ~ 10 dB
400 ~ 10 dB
500 ~ 23 dB
600 ~ 15 dB
Results
Budget and Timeline
Proposed Timeline
Actual Timeline
Sep-07 Oct-07 Nov-07 Dec-07
Develop Concept For Product
Place Orders
Begin Development Cycle
Functional Spec
Design Spec
Assembly of Modules
Develop Embedded Software
Debugging
Prototype Modification / Optimization
Final Report
Proposed & Actual Budget
Item Predicted Cost
Actual Cost
Difference
Texas Instrument TMS320C6713 DSK
150 $480 $330
Audio Accessories(Cables, Adaptors, etc)
150 $126 -$24
Pillow 100 $0 -$100
Miscellaneous(Book, Interface, etc)
100 $40 -$60
Total 500 $646 $146
Future Improvements
Future Improvements
Try more algorithms Automatic Gain ControlFaster convergence rate for complex audi
o processingControllable pre-amplifier and output-ampli
fier
Future Improvements – cont.
More suitable equipment Low frequency Omni-directional microphones Low frequency speakers
Perform testing in a controlled environment
Wideband ANC Solution: Multi-channel System!
Conclusion
What was learned
Time management Mike was wrong! “Take what you think and multi
ply it by 3.” …More like by 8
Team workDSP Active Noise CancellationDocumentationIdeas to Product
Conclusion
Target more complex soundsAutomatic Gain ControlStabilitySolutions…
Multi-channel System! Omni-directional Microphones Low frequency speakers More optimization!!
References
[1] American Physical Therapy Association, “Physical Therapy Patient Satisfaction Questionnaire Research Grants”, 2007, http://www.apta.org//AM/Template.cfm?Section=Home
[2] Texas Instruments, “Design of Active Noise Control System with the TMS320 Family, June 1996, http://focus.ti.com/lit/an/spra042/spra042.pdf
[3] Speech Vision Robotics group , “Finite Impulse Response Filters”, http://svr-www.eng.cam.ac.uk/~ajr/SA95/node13.html
[4] TMS320C6713 DSK - Technical Reference. Stafford, TX: Spectrum Digital Inc., 2004. [5] A DSP/BIOS AIC23 Codec Device Driver for the TMS320DM642 EVM, Texas Instrument, Jun
e 2003, http://focus.ti.com/lit/an/spra922/spra922.pdf [6] “Sampling rate” – Wikipedia, September 2007, http://en.wikipedia.org/wiki/Sampling_rate [7] “Understanding Active Noise Cancellation”, Colin N Hansen, 2001 [8] "Headphones." Frontech - Best of Its Kind. 2006. 1 Nov. 2007 <http://www.frontechonline.com
/headphones.html>. [9] "X-540." Logitech. 2007. 1 Nov. 2007 <http://www.logitech.com/index.cfm/speakers_audio/ho
me_pc_speakers/devices/234&cl=ca,en>. [10]“Latex Pillows, Foam Pillows for Head and Neck”, AllergyBuyersClub. 2007 <http://www.allerg
ybuyersclubshopping.com/latex-head-neck-pillows.html> [11] “A Host Port Interface Board to Enhance the TMS320C6713 DSK” Morrow, M.G.; Welch, T.
B.; Wright, C.H.G. May 2006 <http://ieeexplore.ieee.org>.
Acknowledgement
Dr. Andrew Rawicz Wighton Professor for Engineering Development, School of E
ngineering Science, SFU Mr. Mike Sjoerdsma
Lecturer, School of Engineering Science, SFU Mr. Brad Oldham
Teaching Assistant, School of Engineering Science, SFU Ms. Lisette Paris-Shaadi
Teaching Assistant, School of Engineering Science, SFU Dr. Lakshman One
Professor, School of Engineering Science, SFU
Questions?
Technical Presentation
Block Diagram
Secondary Path Estimation
E = fir_out - adaptfir_out; //error signal
adaptfir_out +=(c[i]*dly_adapt[i]); //adaptive filter filter output
c[i] = c[i]+(beta*E*dly_adapt[i]); //update weights of adaptive filter
FXLMS Implementation
A[n] = 0.9999*A[n]+(muA*En*X[n]); //update weights of adaptive FIR
Xp[0] += (w[l]*X[l]);
Y[0] +=(A[i]*X[i])*10000; //adaptive FIR filter output
Leaky Implementation
A[n] = 0.9999*A[n]+(muA*En*X[n]); //update weights of adaptive FIR
Roundoff and quantization error can accumulate and cause coefficients to grow out of the allowed range (overflow)
Results-200Hz
Results-300Hz
Results-400Hz
Results-500Hz
Results-600Hz
Results-400&600Hz