Non-linear echo cancellation
View Patent ↗A two-stage structure for performing non-linear echo cancellation is described in which a first echo canceller is used to attenuate linear echo components of a microphone signal and a second echo canceller is used to attenuate non-linear echo components of the output signal generated by the first echo canceller. One or both of the echo cancellers may be implemented using closed-form solutions, including a closed form solution for a hybrid method in the frequency domain.
1. An echo cancellation system, comprising:
a first echo canceller that is operable to receive a far-end audio signal and a microphone signal, to utilize the far-end audio signal to generate an estimated linear echo signal, and to subtract the estimated linear echo signal from the microphone signal to generate a first estimated near-end audio signal;
a preprocessor operable to apply an adaptive time domain preprocessing function to the far-end audio signal to generate a non-linear echo reference signal; and
a second echo canceller that is operable to receive the non-linear echo reference signal and the first estimated near-end audio signal, to utilize the non-linear echo reference signal to generate an estimated non-linear echo signal, and to subtract the estimated non-linear echo signal from the first estimated near-end audio signal to generate a second estimated near-end audio signal.
2. The system of claim 1 , wherein the first echo canceller is operable to receive a frequency domain representation of the far-end audio signal and a frequency domain representation of the microphone signal and to generate a frequency domain representation of the first estimated near-end audio signal; and
wherein the second echo canceller is operable to receive the frequency domain representation of the first estimated near-end audio signal and a frequency domain representation of the non-linear echo reference signal and to generate a frequency domain representation of the second estimated near-end audio signal.
3. The system of claim 1 , wherein the first echo canceller comprises a filter that filters the far-end audio signal to generate the estimated linear echo signal, the filter being determined from estimated statistics associated with the far-end audio signal and the microphone signal.
4. The system of claim 2 , wherein the first echo canceller generates the estimated linear echo signal by passing each of a plurality of frequency components of the frequency domain representation of the far-end audio signal through a respective one of a plurality of time direction filters, wherein filter taps associated with each time direction filter are determined from estimated statistics associated with the far-end audio signal and the microphone signal.
5. The system of claim 1 , wherein the second echo canceller comprises a filter that filters the non-linear echo reference signal to generate the estimated non-linear echo signal, the filter being determined from estimated statistics associated with the non-linear echo reference signal and the first estimated near-end audio signal.
6. The system of claim 2 , wherein the second echo canceller generates the estimated non-linear echo signal by passing each of a plurality of frequency components of the frequency domain representation of the non-linear echo reference signal through a respective one of a plurality of time direction filters, wherein filter taps associated with each time direction filter are determined from estimated statistics associated with the non-linear echo reference signal and the first estimated near-end audio signal.
7. The system of claim 1 , wherein the preprocessor comprises a memoryless limiter.
8. The system of claim 1 , wherein the preprocessor is operable to apply a steepest descent adaptation of a preprocessing function to generate the non-linear echo reference signal.
9. The system of claim 1 , wherein the first echo canceller is operable to generate the estimated linear echo signal based at least in part on statistics associated with the far-end audio signal and the microphone signal; and
wherein the first echo canceller is further operable to update the statistics at a rate determined at least in accordance with a measure of coherence between the far-end audio signal and the microphone signal.
10. The system of claim 1 , wherein the first echo canceller is operable to:
generate a first estimated linear echo signal based at least in part on statistics associated with the far-end audio signal and the microphone signal that are updated in accordance with an adaptive update rate;
generate a first signal by subtracting the first estimated linear echo signal from the microphone signal;
generate a second estimated linear echo signal based at least in part on statistics associated with the far-end audio signal and the microphone signal that are updated in accordance with a fixed update rate;
generate a second signal by subtracting the second estimated linear echo signal from the microphone signal; and
generate the first estimated near-end audio signal by selecting for each of a plurality of frequency bins a frequency component from among a frequency component of the first signal and a frequency component of the second signal.
11. The system of claim 1 , wherein the first echo canceller is operable to:
generate a first signal by subtracting the estimated linear echo signal from the microphone signal; and
generate the first estimated near-end audio signal by selecting for each of a plurality of frequency bins a frequency component from among a frequency component of the first signal and a frequency component of the microphone signal.
12. The system of claim 1 , wherein the second echo canceller is operable to generate the estimated non-linear echo signal based at least in part on statistics associated with the non-linear echo reference signal and the first estimated near-end audio signal; and
wherein the second echo canceller is further operable to update the statistics at a rate determined at least in accordance with a measure of coherence between the non-linear echo reference signal and the first estimated near-end audio signal.
13. The system of claim 1 , wherein the second echo canceller is operable to:
generate a first estimated non-linear echo signal based at least in part on statistics associated with the non-linear echo reference signal and the first estimated near-end audio signal that are updated in accordance with an adaptive update rate;
generate a first signal by subtracting the first estimated non-linear echo signal from the first estimated near-end audio signal;
generate a second estimated non-linear echo signal based at least in part on statistics associated with the non-linear echo reference signal and the first estimated near-end audio signal that are updated in accordance with a fixed update rate;
generate a second signal by subtracting the second estimated non-linear echo signal from the first estimated near-end audio signal; and
generate the second estimated near-end audio signal by selecting for each of a plurality of frequency bins a frequency component from among a frequency component of the first signal and a frequency component of the second signal.
14. The system of claim 1 , wherein the second echo canceller is operable to:
generate a first signal by subtracting the estimated non-linear echo signal from the first estimated near-end audio signal; and
generate the second estimated near-end audio signal by selecting for each of a plurality of frequency bins a frequency component from among a frequency component of the first signal and a frequency component of the first estimated near-end audio signal.
15. A method for performing echo cancellation, comprising:
receiving a far-end audio signal and a microphone signal;
utilizing the far-end audio signal to generate an estimated linear echo signal;
subtracting the estimated linear echo signal from the microphone signal to generate a first estimated near-end audio signal;
applying a preprocessing function in the time domain to the far-end audio signal to generate a non-linear echo reference signal;
utilizing the non-linear echo reference signal to generate an estimated non-linear echo signal in the frequency domain; and
subtracting the estimated non-linear echo signal from the first estimated near-end audio signal to generate a second estimated near-end audio signal.
16. The method of claim 15 , wherein applying the preprocessing function to the far-end audio signal comprises passing the far-end audio signal through a memoryless limiter.
17. The method of claim 15 , wherein applying the preprocessing function to the far-end audio signal comprises applying an adaptive preprocessing function to the far-end audio signal.
18. The method of claim 17 , wherein applying the adaptive preprocessing function to the far-end audio signal further comprises performing a steepest descent adaptation of the preprocessing function.
19. A computer program product comprising a computer-readable storage device having computer logic recorded thereon, the computer logic being executable by a processing unit to cause the processing unit to perform steps comprising:
receiving a far-end audio signal and a microphone signal;
utilizing the far-end audio signal to generate an estimated linear echo signal;
subtracting the estimated linear echo signal from the microphone signal to generate a first estimated near-end audio signal;
applying an adaptive preprocessing function that includes passing the far-end audio signal through a memoryless limiter to the far-end audio signal to generate a non-linear echo reference signal;
utilizing the non-linear echo reference signal to generate an estimated non-linear echo signal; and
subtracting the estimated non-linear echo signal from the first output signal to generate a second estimated near-end audio signal.
20. The computer program product of claim 19 , wherein applying the adaptive preprocessing function further comprises performing a steepest descent adaptation of the preprocessing function.