Voice-Isolating Wireless Communicator

1. Voice-Isolating Wireless Communicator Alexander Joo, Ryo Kondo, Frank Lam Team 22 ECE 445 Senior Design April 25, 2008

2. Objective Design Components Complete System Successes and Challenges Recommendations Presentation Overview • Objective • Introduction of problem • Proposed solution • Design • Components • Complete System • Successes and Challenges • Recommendations • Objective • Introduction of problem • Proposed solution • Design • Components • Complete System • Successes and Challenges • Recommendations

3. Noise in the environment can affect the ability to communicate effectively Problem http://www.solarnavigator.net/ http://rampbingspoon.com http://mksviews.wordpress.com

4. Objectives Develop an audio filter that can remove background noise while preserving voice. (by at least 10-20 dB) Wirelessly transmit this cleaned signal over unlicensed spectrum. Perform all functions in real/near-real time. (<100 ms)

5. Voice-Isolating Wireless Communicators Remove noise that disrupts communication Solution

6. Objective Design Components Complete System Successes and Challenges Recommendations Presentation Overview • Objective • Design • Original Design • Alterations and Final Design • Components • Complete System • Successes and Challenges • Recommendations • Objective • Design • Original Design • Alterations and Final Design • Components • Complete System • Successes and Challenges • Recommendations

7. Original Design – Block Diagram

8. Original Design – Block Diagram

9. Original Design – Block Diagram

10. New Design – Block Diagram

11. New Design – Block Diagram

12. New Design – Block Diagram

13. Presentation Overview Objective Design Components Complete System Successes and Challenges Recommendations • Objective • Design • Components • Microphone Array/Preamp • DSP/Algorithm • Wireless • Complete System • Successes and Challenges • Recommendations • Objective • Design • Components • Microphone Array/Preamplifier • DSP/Algorithm • Wireless • Complete System • Successes and Challenges • Recommendations

14. Microphone & Pre-Amp

15. OKAY.II EM320 Clip-on Mini Microphones Factors Price Size Directionality Microphone

16. Microphone Array

17. First point of optimization Ensure that voice signal capture by the “noise-only” microphone is kept to a minimum Microphone Placement

18. Microphones need an external voltage to drive the signal “Mic-In” Signal vs “Line-In” Signal Mic-in jacks provide a source to power the microphones Typical voltage levels are very different Mic-in: 10 mV Line-in: 100 mV Preamplifier

19. Preamplifier Circuit Diagram 5 V 47 k 2.4 k Op Amp 10 micro 10 micro 56 k Audio Jack 47 k 10 micro 2.4 k Audio Jack 5 V 47 k 2.4 k 100 micro Op Amp 10 micro 10 micro Audio Jack 56 k 47 k 10 micro 2.4 k

20. Op Amp is a LF 353 Bandwidth = 4 MHz Amplification Determined by the ratio of R2 and R1 Vo = R2/R1 – 1 Theoretical Vo = 22.3 Tested to confirm response to multiple frequencies Specifications

21. Preamplifier Circuit Diagram 5 V 47 k 2.4 k Op Amp 10 micro 10 micro 56 k Audio Jack <- R2 47 k 10 micro <- R1 2.4 k Audio Jack 5 V 47 k 2.4 k 100 micro Op Amp 10 micro 10 micro Audio Jack <- R2 56 k 47 k 10 micro <- R1 2.4 k

22. Op Amp is a LF 353 Bandwidth = 4 MHz Amplification Determined by the ratio of R2 and R1 Vo = R2/R1 – 1 Theoretical Vo = 22.3 Tested to confirm response to multiple frequencies Specifications

23. Test Case – 700 Hz Input Signal 24 mV P-P Output Signal 500 mV P-P

24. Test Case – 2000 Hz Input Signal 18 mV P-P Output Signal 380 mV P-P

25. Preamplifier PCB

26. TI TMS320C6713 DSP Chip

27. DSP – The Board TI TMS320C6713 DSK – Development kit Speaker Line out Line in Mic in TI DSP Chip

28. DSP – Programming Process • G – “Visual programming language” • C/Assembly • DSP compiled code

29. DSP – Why Labview? Avoid issues with C/Assembly syntax Abstract away low-level details Already implemented DSP functions VI-system provides easy-to-control interface Seamless integration with Code-composer and DSP board.

30. DSP – Original proof of concept Noise signal input Desired signal input Noise subtracted Signals added

32. DSP – Why not subtraction? Proof of concept shows that Labview can be used for audio manipulation… but… Exact timing (no delay) required for both desired signal and noise. Real-world “noise” signals have unpredictable delays.

33. DSP – Least Mean Squares (LMS) Adaptive Filter Filter ĥ(n) attempts to model h(n) using only: x(n) – reference signal (noise + signal) d(n) – noise e(n) – modeled noise contaminating signal

34. DSP – Least Mean Squares (LMS) Adaptive Filter Does so by: Minimizing the cost function Begins with arbitrary values for weights Updating the weights of the filter coefficients With a minimum step size μ.

35. DSP – LMS filter efficiency Depends on: Sampling rate Filter order Convergence value(step-size) Shetty, Kiran Kumar – Least Mean Squares Description, Florida State University, 2004

36. DSP – Filter implementation

37. DSP – Sampling rate and filter order Maximum sampling rate and filter order constrained by maximum processing speed of DSP Board.

38. DSP – Convergence value (step size) • Humans have a higher tolerance to noise with speech. • Optimal value 0.1 – 0.5.

39. DSP – Filter results Scale: 5 kHz window Center: 2.5 kHz Signal: 1 kHz sine Vertical scale: 10 dBv/ Noise alone (600 Hz sine) Noise alone (white noise) Signal alone

40. DSP – Filter results Sine wave White noise

41. DSP – Filter results Music – Jimmy Eat World – A Praise Chorus Music alone Signal + Music w/o filter Signal + Music w/filter

42. DSP – Filter frequency response 40 Hz

43. DSP – Filter frequency response 200 Hz

44. DSP – Filter frequency response 400 Hz

45. DSP – Filter frequency response 600 Hz

46. DSP – Filter frequency response 800 Hz

47. DSP – Filter frequency response 1200 Hz

48. DSP – Filter frequency response 1600 Hz

49. DSP – Filter frequency response 2000 Hz

50. DSP – Filter frequency response 2400 Hz