Systems and methods for adaptive echo cancellation
View Patent ↗An echo canceller comprises foreground and background filters. The background filter locates and confirms peaks, defines active regions centered about the confirmed peaks and updates coefficients of the foreground filter when the background filter is more effective than the foreground filter. A method for improved adaptive echo cancellation comprises configuring a foreground filter, calculating an echo-return loss responsive to the foreground filter, identifying the location of one or more peaks repetitively in a block of data in a background filter, confirming one or more identified peaks over a defined number of blocks before defining an active region equally about the identified peaks in the background filter, filtering the sparse impulse response responsive to the active regions, calculating an echo-return loss responsive to the background filter, and updating the foreground filter when the background filter includes a more effective set of filter coefficients.
1. A method for adaptive echo cancellation, comprising:
configuring an adaptive foreground filter to apply to a block of data including a sparse impulse response to generate a foreground filtered response;
calculating an echo return loss responsive to the foreground filtered response;
identifying the location of one or more peaks repetitively in the block of data in a background filter;
confirming one or more identified peaks over a defined number of blocks before defining an active region equally about the identified peaks in the background filter;
filtering the sparse impulse response responsive to at least one active region,
calculating an echo return loss responsive to the background filtered sparse impulse response;
comparing the second echo return loss responsive to the background filter with the first echo return loss responsive to the foreground filter; and
when a peak is confirmed and when the second echo return loss exceeds the first echo return loss, replacing coefficients from the adaptive foreground filter by the coefficients from the background filter to create a modified foreground filter.
2. The method of claim 1 , wherein identifying the location of one or more peaks repetitively is responsive to a sub-band filtered and decimated signal.
3. The method of claim 1 , wherein confirming one or more identified peaks over a defined number of blocks comprises applying a modified improved proportionate normalized least-mean square (IPNLMS) algorithm.
4. The method of claim 3 , wherein the modified IPNLMS algorithm comprises populating an array of sampled peak values.
5. The method of claim 3 , wherein the modified IPNLMS algorithm comprises calculating a peak vector.
6. The method of claim 5 , wherein the modified IPNLMS algorithm comprises calculating the frequency of stored indices in the peak vector.
7. The method of claim 1 , wherein defining an active region equally about identified peaks comprises analyzing and adjusting the width of each respective active region to prevent overlapping active regions.
8. An adaptive echo canceller embodied in a non-transitory computer-readable medium that when executed by one or more processors filters sampled data, the adaptive echo canceller, comprising:
a foreground filter configured with coefficients expected to remove echo induced peaks in a sparse impulse response and calculate a first echo return loss value; and
a background filter configured to:
bandpass filter and decimate the sparse impulse response;
repetitively identify the location of one or more peaks in the sparse impulse response;
confirm one or more identified peaks over a defined number of blocks of the sparse impulse response before defining an active region equally about the one or more identified peaks such that when a peak is confirmed an adaptive normalized least-mean square algorithm is applied to filter the sparse impulse response; and
calculate a second echo return loss value responsive to the filtered sparse impulse response and
when a peak is confirmed and when the first and second echo return losses are compared, and when the second echo return loss value exceeds the first echo return loss value, coefficients in the foreground filter are replaced with one or more coefficients from the background filter to modify the foreground filter.
9. The adaptive echo canceller of claim 8 , wherein the background filter confirms one or more identified peaks by applying a modified improved proportionate normalized least-mean square (IPNLMS) algorithm.
10. The adaptive echo canceller of claim 9 , wherein the modified IPNLMS algorithm comprises populating an array of sampled peak values.
11. The adaptive echo canceller of claim 9 , wherein the modified IPNLMS algorithm comprises calculating a peak vector.
12. The adaptive echo canceller of claim 11 , wherein the modified IPNLMS algorithm comprises determining the frequency with which indices are above a threshold value in the peak vector.
13. The adaptive echo canceller of claim 8 , wherein the background filter defines an active region equally about identified peaks by analyzing and adjusting the width of each respective active region to prevent overlapping active regions.
14. The adaptive echo canceller of claim 8 , wherein the foreground filter is configured to operate in one of an adaptive mode and a fixed mode and applied to the sampled signal, the background filter is configured to operate in one of first, second, and third operational modes.
15. An adaptive echo canceller, comprising:
means for determining the effectiveness of a foreground filter at removing one or more echo induced peaks in a sparse impulse response;
means for repetitively identifying the location of a user defined number of peaks responsive to a bandpass filtered and decimated representation of the sparse impulse response;
means for storing a user defined number of peak detection vectors responsive to peaks in the representation of the sparse impulse response;
means for confirming the location of a peak responsive to the frequency of a respective element within the peak detection vectors; means for defining one or more active regions centered about respective confirmed peaks;
means for background filtering the representation of the sparse impulse response responsive to the one or more active regions;
means for determining the effectiveness of the means for background filtering at removing peaks in the representation of the sparse impulse response; and
means for comparing the effectiveness of the foreground filter to the effectiveness of the means for background filtering, wherein when the means for background filtering is more effective than the foreground filter, one or more coefficients in the foreground filter are replaced with coefficients associated with the means for background filtering to create an improved foreground filter.
16. The adaptive echo canceller of claim 15 , wherein the means for confirming the location of a peak comprises a modified improved proportionate normalized least-mean square (IPNLMS) algorithm.
17. The adaptive echo canceller of claim 16 , wherein the modified IPNLMS algorithm comprises populating an array responsive to sampled peak values.
18. The adaptive echo canceller of claim 16 , wherein the modified IPNLMS algorithm comprises calculating a peak vector.
19. The adaptive echo canceller of claim 18 , wherein the modified IPNLMS algorithm comprises determining the frequency with which indices are above a threshold value in the peak vector.
20. The adaptive echo canceller of claim 15 , wherein the means for defining active regions adjusts the width of each respective active region to prevent overlapping between adjacent active regions.