Acoustic echo suppression

• US 8,325,909 B2
• Filed: 06/25/2008
• Issued: 12/04/2012
• Est. Priority Date: 06/25/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Tashev, Ivan J., Acero, Alejandro, Madhu, Nilesh
• Primary Examiner(s): Tsang, Fan
• Assistant Examiner(s): LYTLE, JEFFREY P
• Application Number:
  US12/145,579
• Publication Number:
  US 20090323924A1
• Time in Patent Office: 1,623 Days
• Field of Search: None
• US Class Current: 379/406.05
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23

Abstract

Sound signals captured by a microphone are adjusted to provide improved sound quality. More particularly, an Acoustic Echo Reduction system which performs a first stage of echo reduction (e.g., acoustic echo cancellation) on a received signal is configured to perform a second stage of echo reduction (e.g., acoustic echo suppression) by segmenting the received signal into a plurality of frequency bins respectively comprised within a number of frames (e.g., 0.3 s to 0.5 s sound signal segments) for a given block. Data comprised within respective frequency bins is modeled according to a probability density function (e.g., Gaussian distribution). The probability of whether respective frequency bins comprise predominantly near-end signal or predominantly residual echo is calculated. The output of the acoustic echo suppression is computed as a product of the content of a frequency bin in a frame and the probability the frequency bin in a frame comprises predominantly near-end signal, thereby making near-end signals more prominent than residual echoes.

Claims

1. A method for acoustic echo suppression comprising:

temporally segmenting a sound signal into segmented sound signals stored within a plurality a frequency bins according to frequency and time of capture, the sound signal comprising a near-end signal and a residual echo;

modeling segmented sound signals stored in respective frequency bins as a probability density function to determine a probability that respective frequency bins comprise predominantly near-end signal or predominantly residual echo, the modeling for respective frequency bins comprising;

forming an unknown parameter vector comprising unknown parameters characterizing the probability density function; and

evaluating the unknown parameter vector using a statistical approach to determine the unknown parameters; and

computing an acoustic echo suppression output having a suppressed residual echo, the acoustic echo suppression output proportional to products of segmented sound signals and probabilities associated with respective frequency bins.

2. The method of claim 1, the probability density function comprising a Gaussian distribution.

3. The method of claim 1, the statistical approach comprising an Expectation Maximization algorithm.

4. The method of claim 1, comprising:

incorporating the unknown parameter vector into a log likelihood function.

5. The method of claim 4, the log likelihood function comprising a probability that a given signal output from a speaker will be received at an acoustic echo suppressor configured to perform the method for acoustic echo suppression.

6. The method of claim 1, comprising:

performing acoustic echo cancellation (AEC) prior to temporally segmenting the sound signal.

7. The method of claim 1, comprising:

converting the sound signal to a frequency domain prior to temporally segmenting the sound signal.

8. The method of claim 1, comprising:

segmenting the sound signal into 0.01 s to 0.05 s segmented sound signals.

9. A system for acoustic echo reduction, comprising:

an acoustic echo suppressor configured to;

temporally segment a sound signal into segmented sound signals stored within a plurality a frequency bins according to frequency and time of capture, the sound signal comprising a near-end signal and a residual echo;

model segmented sound signals stored in respective frequency bins as a probability density function to determine a probability that respective frequency bins comprise predominantly near-end signal or predominantly residual echo, the modeling for respective frequency bins comprising;

forming an unknown parameter vector comprising unknown parameters characterizing the probability density function; and

evaluating the unknown parameter vector using a statistical approach to determine the unknown parameters; and

compute an acoustic echo suppression output having a suppressed residual echo, the acoustic echo suppression output proportional to products of segmented sound signals and probabilities associated with respective frequency bins.

10. The system of claim 9, comprising:

an acoustic echo canceller configured to perform acoustic echo cancellation (AEC) prior to temporally segmenting the sound signal.

11. The system of claim 9, the probability density function comprising a Gaussian distribution.

12. The system of claim 9, the statistical approach comprising an Expectation Maximization algorithm.

13. The system of claim 9, the acoustic echo suppressor configured to:

incorporate the unknown parameter vector into a log likelihood function.

14. The system of claim 13, the log likelihood function comprising a probability that a given signal output from a speaker will be received at the acoustic echo suppressor.

15. The system of claim 9, the acoustic echo suppressor configured to:

convert the sound signal to a frequency domain prior to temporally segmenting the sound signal.

16. The system of claim 9, the acoustic echo suppressor configured to:

segment the sound signal into 0.01 s to 0.05 s segmented sound signals.

17. A computer storage device comprising computer executable instructions that when executed via a processing unit perform a method for acoustic echo suppression, comprising:

temporally segmenting a sound signal into segmented sound signals stored within a plurality a frequency bins according to frequency and time of capture, the sound signal comprising a near-end signal and a residual echo;

modeling segmented sound signals stored in respective frequency bins as a probability density function to determine a probability that respective frequency bins comprise predominantly near-end signal or predominantly residual echo, the modeling for respective frequency bins comprising;

forming an unknown parameter vector comprising unknown parameters characterizing the probability density function; and

evaluating the unknown parameter vector using a statistical approach to determine the unknown parameters; and

computing an acoustic echo suppression output having a suppressed residual echo, the acoustic echo suppression output proportional to products of segmented sound signals and probabilities associated with respective frequency bins.

18. The method of claim 17, the probability density function comprising a Gaussian distribution.

19. The method of claim 17, the statistical approach comprising an Expectation Maximization algorithm.

20. The method of claim 17, comprising:

incorporating the unknown parameter vector into a log likelihood function, the log likelihood function comprising a probability that a given signal output from a speaker will be received at an acoustic echo suppressor configured to perform the method for acoustic echo suppression.