Microphone occlusion detector

• US 9,100,756 B2
• Filed: 12/14/2012
• Issued: 08/04/2015
• Est. Priority Date: 06/08/2012
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Dusan, Sorin V., Yeh, David T., Lindahl, Aram M., Kanaris, Alexander
• Primary Examiner(s): Bernardi, Brenda
• Application Number:
  US13/715,422
• Publication Number:
  US 20130329895A1
• Time in Patent Office: 963 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: H04R 2410/05 Noise reduction with a separate noise microphone
• CPC Class Codes: H04R 2499/11 Transducers incorporated or for use in hand-held devices, e.g. mobile phones, PDA's, camera's
• CPC Class Codes: H04R 29/00 Monitoring arrangements; Testing arrangements for hearing aids H04R25/30; detection of loudspeaker connection H04R5/04; sound-field adaptation dependent on speaker detection H04S7/308
• CPC Class Codes: H04R 29/004 for microphones H04R29/007 takes precedence
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407

Abstract

Digital signal processing microphone occlusion detection is described that can be used with a noise suppression system that uses two types of noise estimators, including a more aggressive one based on two audio signals (such as for non-stationary noises) and a less aggressive one based on one audio signal (such as for stationary noises). Decisions are made on how to select or combine the outputs of the noise estimators into a usable noise estimate, based on an occlusion function. The occlusion detection may alternatively be used to trigger an alert to users of multi-microphone audio processing systems, such as smart phones, headsets, laptops and tablet computers. Other embodiments are also described and claimed.

Claims

1. An electronic system for audio noise processing and for noise reduction, using a plurality of microphones, comprising:

a first noise estimator to process a first audio signal from a first one of the microphones, and generate a first noise estimate;

a second noise estimator to process the first audio signal, and a second audio signal from a second one of the microphones, in parallel with the first noise estimator, and generate a second noise estimate;

a combiner-selector to receive the first and second noise estimates, and to generate an output noise estimate using the first and second noise estimates; and

a microphone occlusion detector to process the first and second audio signals including to band pass filter the first and second audio signals and to generate a microphone occlusion signal using the processed first and second audio signals, wherein the microphone occlusion signal represents a measure of how severely or how likely it is that one of the microphones is occluded, and

wherein the combiner-selector is to generate its output noise estimate also based on the occlusion signal.

2. The system of claim 1 wherein the combiner-selector selects the first noise estimate for its output noise estimate, and not the second noise estimate, when the occlusion signal indicates that the second one of the microphones is substantially occluded.

3. The system of claim 1 wherein the occlusion detector computes a power or energy ratio (PR) or a magnitude ratio (MR) of band pass filtered versions of the first and second audio signals, and evaluates an occlusion function at the computed PR or MR.

4. The system of claim 3 wherein the occlusion detector is to band pass filter the first and second audio signals over a pass band of about 2000Hz-4000Hz.

5. The system of claim 3 wherein when the PR or MR is greater than a threshold the occlusion function has a fixed value indicating substantial occlusion, and when the PR or MR is less than the threshold the occlusion function has a different fixed value that indicates no substantial occlusion.

6. The system of claim 3 wherein the PR is computed using the formula

PR=pow1t−

pow2t

where

pow1t=10*log 10{[summation of frame_mic1(i)*frame_mic1(i)]/N},

pow2t=10*log 10{[summation of frame_mic2(i)*frame_mic2(i)]/N}

and where frame_mic1 and frame_mic2 include samples from i=1 to i=N of band pass filtered versions of the signals from the first and second microphones, or it is computed in the frequency domain by summation over a range of frequency bins.

7. The system of claim 6 wherein the PR is computed in dB.

8. The system of claim 1 wherein the occlusion detector further computes one or more of the following and uses them to generate the occlusion signal:

absolute power of the second audio signal over a given time frame and over a range of frequency bins within the given time frame;

the output noise estimate per frequency bin; and

a VAD decision per frequency bin.

9. The system of claim 6 wherein the occlusion function is proportional to a logistic function C/(1+A*exp(−

B*PR)) where A, B and C are scalar coefficients that define the slope, position and final magnitude of the logistic function.

10. A microphone occlusion detector comprising:

means for processing first and second audio signals that are from first and second microphones, respectively, including means for band pass filtering the signals; and

means for evaluating a microphone occlusion function that represents a measure of how severely or how likely it is that the second microphone is occluded, using the processed first and second audio signals.

11. The occlusion detector of claim 10 wherein the processing means computes a power ratio (PR) or magnitude ratio (MR) of band pass filtered versions of the first and second audio signals in a pass band of about 2000Hz-4000Hz, and the occlusion function is a function of the PR or MR.

12. The occlusion detector of claim 11 wherein the occlusion function takes on a high value that indicates occlusion or greater occlusion when the PR or MR is greater than a threshold, and the occlusion function takes on a low value that indicates no occlusion or lesser occlusion when the PR or MR is less than the threshold.

13. The occlusion detector of claim 11 wherein the power ratio is computed using the formula

PR=pow1t−

pow2t

where

pow1t=10*log10{[summation of frame_mic1(i)*frame_mic1(i)]/N},

pow2t=10*log 10{[summation of frame_mic2(i)*frame_mic2(i)]/N}

where frame_mic1 and frame_mic2 include samples from i=1 to i=N of band pass filtered versions of audio signals from the first microphone and the second microphone, or is computed in the frequency domain by summing power spectrum bins from start to stop where start and stop are the frequency bins that define the boundaries of the band pass filter pass band.

14. The occlusion detector of claim 10 wherein the occlusion function is proportional to a logistic function C/(1+A*exp(−

B*PR)) where A, B and C are scalar coefficients that define the slope, position and final magnitude of the logistic function.

15. The occlusion detector of claim 14 further comprising:

smoothing means for smoothing the logistic function.

16. A microphone occlusion detector that uses multiple occlusion component signals, comprising:

means for processing first and second audio signals that are from first and second microphones, respectively, including means for band pass filtering the signals;

means for evaluating a first occlusion component function that represents a measure of how severely or how likely it is that the second microphone is occluded during speech activity, using the band pass filtered first and second audio signals;

means for evaluating a second occlusion component function that represents a measure of how severely or how likely it is that the second microphone is occluded during no speech activity, using the band pass filtered first and second audio signals; and

means for selecting between the first and second occlusion component functions,

wherein in a no speech condition, the second component function is selected but not the first component function, and

in a no speech condition where the level of background noise is at a high level, the first component function is selected but not the second compound function.

17. The occlusion detector of claim 16 wherein the audio signal processing means comprises a voice activity detector that indicates said no speech condition.

18. The occlusion detector of claim 16 wherein the audio signal processing means comprises means for computing absolute power of the second audio signal to indicate the level of background noise is at a low level.

19. The occlusion detector of claim 16 wherein the audio signal processing means comprises a background noise estimator.

20. The occlusion detector of claim 16 wherein each of the first and second occlusion component functions is a logistic function, each being a function of a power ratio (PR) or magnitude ratio (MR) of the first and second audio signals.

21. The occlusion detector of claim 20 wherein an inflection point of the first component function is at a lower PR or MR value than that of the second component function.

22. A method for detecting occlusion of a microphone, comprising:

processing band pass filtered versions of first and second audio signals that are from first and second microphones, respectively, to compute a power ratio (PR) or a magnitude ratio (MR) of the band pass filtered first and second signals; and

evaluating an occlusion function, being a measure of how occluded the second microphone is, as a function of PR or MR, wherein the occlusion function is one of

a) a curve that abruptly indicates substantial occlusion when the PR or MR is greater than a threshold, and abruptly indicates no substantial occlusion when the PR or MR is smaller than the threshold,

b) a curve that gradually indicates increasing occlusion when the PR or MR is greater than a threshold, and abruptly indicates no substantial occlusion when the PR or MR is smaller than the threshold, and

c) a logistic function.

23. The method of claim 22 wherein the occlusion function is the logistic function, the method further comprising smoothing the logistic function.

24. The method of claim 22 wherein the logistic function is smoothed using an exponential filter.

25. The method of claim 22 further comprising:

generating a noise estimate from the first audio signal and not the second audio signal, responsive to the evaluated occlusion function indicating more occlusion or occlusion present; and

generating the noise estimate from both the first and second audio signals responsive to the evaluated occlusion function indicating less occlusion or occlusion absent.