Method, apparatus and articles incorporating a step size control technique for echo signal cancellation

1. A method for cancelling an echo signal in a bidirectional communications link comprising:

computing step sizes using power spectral density function of an error signal and a power spectral density function of a far end signal using the equation:

μ (n) ( j )=μ 1 [R XX (n) ( j )+ NR EE (n) ( j )] −1

wherein μ 1 is a value less than 1/N, N being a number of sub filters used in a partitioned block frequency domain adaptive filter (PBFDAF), n is a block index, R XX is the power spectral density function of the far end signal, R EE is the power spectral density function of the error signal, and j is an index to a frequency-bin; and

applying the computed step sizes to the N sub filters in the PBFDAF to cancel an echo signal.

2. The method of claim 1 , wherein computing the power spectral density function of the far end signal comprises:

receiving the far end signal;

forming far end blocks of samples from the received far end signal;

performing an FFT on the far end blocks of samples and obtaining a frequency domain far end signal; and

computing power spectral density function of the far end signal using the obtained frequency domain far end signal.

3. The method of claim 2 , wherein computing the power spectral density function of the error signal comprises:

estimating echo signal by the PBFDAF from the far end signal;

obtaining time domain samples of the estimated echo signal;

subtracting time domain samples of the estimated echo signal from the microphone signal and outputting a time domain-error signal;

performing FFT on the time domain-error signal for obtaining a frequency domain-error signal; and

computing power spectral density function of the error signal using the obtained frequency domain error signal.

4. The method of claim 3 , wherein applying the step sizes comprises:

multiplying the computed step sizes with the frequency domain error signal and the frequency domain far end signal to obtain a correction coefficient; and

obtaining adapted PBFDAF coefficients by adding the correction coefficient to PBFDAF coefficients associated with adaptive filters in the PBFDAF.

5. The method of claim 4 , wherein obtaining the time domain samples of the estimated echo signal comprises:

multiplying the adapted PBFDAF coefficients with the frequency domain far end signal to output a frequency domain estimated echo signal; and

performing IFFT on the frequency domain estimated echo signal to obtain the time domain samples of the estimated echo signal.

6. The method of claim 5 , further comprising: receiving the microphone signal; and

forming current near end blocks of samples from the microphone signal.

7. The method of claim 1 , further comprising:

repeating the computing and applying steps for a next far end and near end blocks of samples.

8. A method of echo signal cancellation in a bidirectional communication link comprising:

computing step sizes using power spectral density functions of an error signal and a power spectral density function of a far end signal using the equation:

μ (n) ( j )=μ 1 [R XX (n) ( j )+ NR EE (n) ( j )] −1

wherein μ 1 is a value Less than 1/N, N being a number of sub filters used in a partitioned block frequency domain adaptive filter (PBFDAF), n is a block index, R XX is the power spectral density function of the far end signal, R EE is the power spectral density function of the error signal, and j is an index to a frequency-bin;

applying the computed step sizes to the N sub filters in the PBFDAF; and

repeating above steps until the echo signal is canceled.

9. The method of claim 8 , wherein the power spectral density function of the far end signal R XX is computed using the equation:

R XX (n) ( j )=λ R XX (n−1) ( j )+(1−λ)| X 0 (n) ( j )| 2

wherein λ is a smoothing constant having a chosen value in the range of about 0 and 1.

10. The method of claim 9 , wherein the power spectral density functions of the error signal R EE is computed using the equation:

R EE (n) ( j )=λ R EE (n−1)( j )+(1−λ)| E (n) ( j )| 2

wherein λ is a smoothing constant having a chosen value in the range of about 0 and 1.

11. A apparatus for echo signal cancellation in a bidirectional communication link, comprising:

an input module to receive a far end signal, wherein the input module to form a far end blocks of samples from the far end speech signal, and wherein the input module to further receive a MIC signal including a near end signal, an echo signal, and a background noise;

a step size processor for computing step sizes using the equation:

μ (n) ( j )=μ 1 [R XX (n) ( j )+ NR EE (n) ( j )] 31 1

wherein μ 1 is chosen a value less than 1/N, N is a number of sub filters, n is a block index, R XX is a power spectral density function of the far end signal, R EE is a power spectral density function of the error signal, and j is an index to a frequency-bin;

a PBFDAF module with N sub filters coupled to the step size processor and input module to receive the computed step sizes and use the step sizes to update PBFDAF coefficients associated with adaptive filters in the PBFDAF to substantially remove the echo signal from the microphone signal; and

an output module coupled to the PBFDAF module to output a substantially pure near end speech signal.

12. The apparatus of claim 11 , wherein the PBFDAF module multiplies the computed step sizes with error signal and the far end signal to obtain correction coefficients, and wherein the PBFDAF module adds the obtained correction coefficients to the associated PBFDAF coefficients.

13. The apparatus of claim 12 , wherein the PBFDAF module subtracts time domain samples of an estimated echo signal by the PBFDAF module from a current block of samples obtained from a near end microphone signal and outputs a time domain-error signal, wherein the PBFDAF module performs an FFT on the time domain-error signal for obtaining a frequency domain error signal, wherein the PBFDAF module obtains far end signal from a current block of far end samples, and wherein the step size processor computes power spectral density functions of the error and far end signals using the obtained error signal and the far end speech signals obtained from the current block of far end samples.

14. An echo canceller for a bidirectional communication link comprising:

a storage medium having instructions that, when executed by a computing platform, result in execution of a method comprising:

computing step sizes using power spectral density function of an error signal and a power spectral density function of a far end signal using the equation:

μ (n) ( j )−μ 1 [R XX (n) ( j )+ NR EE (n) ( j )] −1

wherein μ 1 is a value less than 1/N, N being a number of sub filters used in a partitioned block frequency domain adaptive filter (PBFDAF), n is a block index, R XX is the power spectral density function of the far end signal, R EE is the power spectral density function of the error signal, and j is an index to a frequency-bin; and

applying the computed step sizes to the N sub filters in the PBFDAF to cancel an echo signal.

15. The echo canceller for a bidirectional communications link of claim 14 , wherein computing the power spectral density function of the far end signal comprises:

receiving the far end signal;

forming far end blocks of samples from the received far end signal;

performing an FFT on the far end blocks of samples and obtaining a frequency domain far end signal; and

computing power spectral density function of the far end signal using the obtained frequency domain far end signal.

16. The echo canceller for a bidirectional communications link of claim 15 , wherein computing the power spectral density function of the error signal comprises:

estimating echo signal by the PBFDAF from the far end signal;

obtaining time domain samples of the estimated echo signal;

subtracting time domain samples of the estimated echo signal from the microphone signal and outputting a time domain-error signal;

performing FFT on the time domain-error signal for obtaining a frequency domain-error signal; and

computing power spectral density function of the error signal using the obtained frequency domain error signal.

17. The echo canceller for a bidirectional communications link of claim 16 , wherein applying the step sizes comprises:

multiplying the computed step sizes with the frequency domain error signal and the frequency domain far end signal to obtain a correction coefficient; and

obtaining adapted PBFDAF coefficients by adding the correction coefficient to PBFDAF coefficients associated with adaptive filters in the PBFDAF.

18. The echo canceller for a bidirectional communications link of claim 17 , wherein obtaining the time domain samples of the estimated echo signal comprises:

multiplying the adapted PBFDAF coefficients with the frequency domain far end signal to output a frequency domain estimated echo signal; and

performing IFFT on the frequency domain estimated echo signal to obtain the time domain samples of the estimated echo signal.

19. The echo canceller for a bidirectional communications link of claim 18 , further comprising:

receiving the microphone signal; and

forming current near end blocks of samples from the microphone signal.

20. The echo canceller for a bidirectional communications link of claim 14 , further comprising:

repeating the computing and applying steps for a next set of far end and near end blocks of samples.