System and Method for Acoustic Echo Removal (AER)

• US 20080159551A1
• Filed: 03/30/2007
• Published: 07/03/2008
• Est. Priority Date: 12/28/2006
• Status: Abandoned Application

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Gupta, Puneet, Harley, Thomas Randall, Kosanovic, Bogdan
• Application Number:
  US11/693,850
• Publication Number:
  US 20080159551A1
• Time in Patent Office: Days
• US Class Current: 381/66
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23

Abstract

One embodiment of the present invention includes an acoustic echo removal system. A transmit signal and a receive signal each having high and low frequency portions, are propagated between a microphone and at least one voice processor. A first acoustic echo removal portion determines and provides a first variable attenuation gain to the low-frequency portion of the transmit signal at a first sample frequency and provides a second variable attenuation gain to the low-frequency portion of the receive signal at the first sample frequency. A second acoustic echo removal portion provides the first variable attenuation gain to the high-frequency portion of the transmit signal at a second sample frequency and provides the second variable attenuation gain to both the high-frequency portion of the receive signal and a copy of the low-frequency portion of the receive signal at the second sample frequency.

Claims

1. An acoustic echo removal system comprising:

a transmit path configured to propagate a transmit signal between a microphone and at least one voice processor, the transmit signal comprising a high-frequency portion and a low-frequency portion;

a receive path configured to propagate a receive signal between the at least one voice processor and a speaker, the receive signal comprising a high-frequency portion and a low-frequency portion;

a first acoustic echo removal portion configured to determine a first variable attenuation gain and to provide the first variable attenuation gain to the low-frequency portion of the transmit signal at a first sample frequency and to provide a second variable attenuation gain to the low-frequency portion of the receive signal at the first sample frequency; and

a second acoustic echo removal portion configured to provide the first variable attenuation gain to the high-frequency portion of the transmit signal at a second sample frequency and to provide the second variable attenuation gain to both the high-frequency portion of the receive signal and a copy of the low-frequency portion of the receive signal at the second sample frequency, the second sample frequency being greater than the first sample frequency.

2. The system of claim 1, wherein the low-frequency portion of both the transmit signal and the receive signal has a frequency that is less than or equal to 3400 Hz, and wherein the first sample frequency is approximately 8 kHz and the second sample frequency is approximately 16 kHz.

3. The system of claim 1, wherein the transmit path comprises a transmit-path bandsplitter configured to input the transmit signal and to output the low-frequency portion and the high-frequency portion of the transmit signal, and wherein the receive path comprises a receive-path bandsplitter configured to input the receive signal and to output the low-frequency portion, the copy of the low frequency portion, and the high-frequency portion of the receive signal.

4. The system of claim 3, wherein the transmit-path bandsplitter comprises a low-pass filter configured to generate the low-frequency portion of the transmit signal and a subtractor configured to subtract the low-frequency portion of the transmit signal from the transmit signal to generate the high-frequency portion of the transmit signal, and wherein the receive-path bandsplitter comprises a low-pass filter configured to generate the low-frequency portion of the receive signal and a subtractor configured to subtract the low-frequency portion of the receive signal from the receive signal to generate the high-frequency portion of the receive signal.

5. The system of claim 1, wherein the receive path further comprises an adder configured to add the high-frequency portion and the copy of the low-frequency portion of the receive signal to generate an attenuated receive signal, and wherein the transmit path further comprises an upsampler configured to increase a sample frequency associated with the low-frequency portion of the transmit signal from the first sample frequency to the second sample frequency, and an adder configured to add the high-frequency portion and the low-frequency portion of the transmit signal to generate an attenuated transmit signal.

6. The system of claim 5, wherein the high-frequency portion of the transmit signal comprises a compensation component, the compensation component comprising distortion associated with at least one of normalization, downsampling, upsampling, and low-pass filtering of the low-frequency portion of the transmit signal, such that the attenuated transmit signal is a substantially identical reconstruction of the transmit signal.

7. The system of claim 5, wherein the first acoustic echo removal portion comprises an adaptive filter acoustic echo canceller in the transmit path configured to receive the attenuated receive signal and to subtract linearly predicted acoustic echo from the low-frequency portion of the transmit signal based on the attenuated receive signal.

8. The system of claim 1, wherein the acoustic echo removal system is switchable between a wide-band mode and a low-band mode, such that the second acoustic echo removal portion is configured to provide the first variable attenuation gain to the high-frequency portion of the transmit signal at the second sample frequency and to provide the second variable attenuation gain to both the high-frequency portion of the receive signal and the copy of the low-frequency portion of the receive signal at the second sample frequency only in the wide-band mode.

9. The system of claim 1, wherein the first acoustic echo removal portion, upon applying the first variable attenuation gain to the low-frequency portion of the transmit signal, provides an output of the low-frequency portion of the transmit signal at the first sample frequency directly to the at least one voice processor.

10. The system of claim 1, wherein the first acoustic echo removal portion comprises a dominant path decision block that is configured to determine which of the receive signal and the transmit signal is a dominant signal based on an associated signal strength, and wherein the first variable attenuation gain increases and the second variable attenuation gain decreases gradually over a predetermined time upon the receive signal being the dominant signal, the second variable attenuation gain increases and the first variable attenuation gain decreases gradually over the predetermined time upon the transmit signal being the dominant signal, and the high-frequency portion of one of the receive signal and the transmit signal is gradually completely attenuated over the predetermined time upon the other one of the receive signal and the transmit signal becoming the dominant signal.

11. The system of claim 1, wherein each of the transmit path and the receive path comprise a saturation detector operative, upon detecting saturation of the low-frequency portion of the respective transmit signal and receive signal based on a transient overflow, to command the second acoustic echo removal portion to increase the respective one of the first variable attenuation gain and the second variable attenuation gain.

12. An acoustic echo removal system comprising:

a receive-path bandsplitter configured to split a receive signal into a high-frequency portion, a low-frequency portion, and a copy of the low-frequency portion;

a receive-path downsampler configured to reduce a sample frequency associated with the low-frequency portion of the receive signal from a first sample frequency to a second sample frequency;

a receive-path portion of a non-linear processor configured to apply a receive attenuation gain to the low-frequency portion of the receive signal at the second sample frequency;

a receive-path attenuator configured to apply the receive attenuation gain to both the high-frequency portion of the receive signal and the copy of the low-frequency portion of the receive signal; and

a receive-path adder configured to add the high-frequency portion and the copy of the low-frequency portion of the receive signal to generate an attenuated receive signal.

13. The system of claim 12, further comprising:

a transmit-path bandsplitter configured to split a transmit signal into a high-frequency portion and a low-frequency portion;

a transmit-path downsampler configured to reduce a sample frequency associated with the low-frequency portion of the transmit signal from the first sample frequency to the second sample frequency;

a transmit-path portion of a non-linear processor configured to apply a transmit attenuation gain to the low-frequency portion of the transmit signal at the second sample frequency;

a transmit-path attenuator configured to apply the transmit attenuation gain to the high-frequency portion of the transmit signal at the first sample frequency;

a transmit-path upsampler configured to increase the sample frequency associated with the low-frequency portion of the transmit signal from the second sample frequency to the first sample frequency; and

a transmit-path adder configured to add the high-frequency portion and the low-frequency portion of the transmit signal to generate an attenuated transmit signal.

14. The system of claim 13, wherein the transmit-path bandsplitter is further configured to subtract a compensation component from the high-frequency portion of the transmit signal, the compensation component comprising distortion associated with at least one of normalization, downsampling, upsampling, and low-pass filtering of the low-frequency portion of the transmit signal, such that the attenuated transmit signal is a substantially identical reconstruction of the transmit signal.

15. The system of claim 13, wherein the transmit-path portion of a non-linear processor, upon applying the transmit attenuation gain to the low-frequency portion of the transmit signal, provides an output of the low-frequency portion of the transmit signal at the second sample frequency directly to at least one voice processor with which the acoustic echo removal system is associated.

16. The system of claim 13, wherein the transmit-path portion of a non-linear processor comprises an adaptive filter acoustic echo canceller configured to receive the attenuated receive signal and to subtract linearly predicted acoustic echo from the low-frequency portion of the transmit signal based on the attenuated receive signal.

17. The system of claim 13, further comprising a dominant path decision block that is configured to determine which of the receive signal and the transmit signal is a dominant signal based on an associated signal strength, such that the other of the receive signal and the transmit signal is a non-dominant signal, the dominant path decision block also being configured to gradually completely attenuate the high-frequency portion of the non-dominant signal over a predetermined time.

18. The system of claim 17, wherein the transmit attenuation gain and the receive attenuation gain are each variable and substantially complementary based on which of the receive signal and the transmit signal is the dominant signal, such that the transmit attenuation gain is increased and the receive attenuation gain is decreased gradually over the predetermined time upon the receive signal being the dominant signal, and the receive attenuation gain is increased and the transmit attenuation gain is decreased gradually over the predetermined time upon the transmit signal being the dominant signal.

19. The system of claim 12, wherein the receive-path bandsplitter comprises a low-pass filter and a subtractor, such that the receive signal is input to both the low-pass flier and the subtractor, the low-pass filter being configured to generate the low-frequency portion of the receive signal and the subtractor being configured to subtract the low-frequency portion of the receive signal from the receive signal to generate the high-frequency portion of the receive signal.

20. The system of claim 12, wherein the low-frequency portion of the receive signal has a frequency that is less than or equal to 3400 Hz, and wherein the first sample frequency is approximately 16 kHz and the second sample frequency is approximately 8 kHz.

21. The system of claim 12, wherein the receive-path downsampler comprises a saturation defector operative, upon detecting saturation of the low-frequency portion of the receive signal based on a transient overflow, to command the receive-path portion of a non-linear processor to increase the receive attenuation gain.

22. A method of removing acoustic echo in a voice communication device, the method comprising:

bandsplitting a transmit signal into a high-frequency portion and a low-frequency portion;

subtracting a compensation component from the high-frequency portion of the transmit signal, the compensation component comprising distortion associated with at least one of normalization, downsampling, upsampling, and low-pass filtering of the low-frequency portion of the transmit signal;

downsampling the low-frequency portion of the transmit signal from a first sample frequency to a second sample frequency;

applying a first variable attenuation gain on the low-frequency portion of the transmit signal at the second sample frequency and on the high-frequency portion of the transmit signal at the first sample frequency;

upsampling the low-frequency portion of the transmit signal from the second sample frequency to the first sample frequency; and

adding the low-frequency portion and the high-frequency portion of the transmit signal to generate an attenuated transmit signal, such that the attenuated transmit signal is a substantially identical reconstruction of the transmit signal.

23. The method of claim 22, further comprising providing the low-frequency portion of the transmit signal at the second sample frequency directly to at least one voice processor associated with the voice communication device.

24. The method of claim 22, further comprising:

bandsplitting a receive signal into a high-frequency portion, a low-frequency portion, and a copy of the low-frequency portion;

downsampling the low-frequency portion of the receive signal from the first sample frequency to the second sample frequency;

applying a second variable attenuation gain on the low-frequency portion of the receive signal at the second sample frequency, and applying the second variable attenuation gain on the high-frequency portion of the receive signal and the copy of the low-frequency portion of the receive signal at the first sample frequency; and

adding the copy of the low-frequency portion and the high-frequency portion of the receive signal to generate an attenuated receive signal.

25. The method of claim 24, further comprising delaying samples of the attenuated receive signal and adaptively filtering the low-frequency portion of the transmit signal at the second sample frequency based on the delayed samples of the attenuated receive signal.

26. The method of claim 25, further comprising:

comparing a first signal strength associated with the low-frequency portion of the receive signal and a second signal strength associated with the low-frequency portion of the transmit signal;

assigning one of the receive signal and the transmit signal as a dominant signal based on the comparison of the first signal strength and the second signal strength;

assigning the other of the receive signal and the transmit signal as a non-dominant signal;

gradually completely attenuating the high-frequency portion of the non-dominant signal over a predetermined time;

decreasing the first variable attenuation gain and increasing the second variable attenuation gain gradually over the predetermined time upon the transmit signal being the dominant signal; and

decreasing the second variable attenuation gain and increasing the first variable attenuation gain gradually over the predetermined time upon the receive signal being the dominant signal.

27. The method of claim 25, further comprising:

detecting saturation associated with downsampling the low-frequency portions of each of the transmit signal, the receive signal, and the delayed samples of the attenuated receive signal based on a transient overflow;

increasing the second variable attenuation gain in response to detecting saturation of the low-frequency portion of the receive signal;

increasing the first variable attenuation gain in response to detecting saturation of the low-frequency portion of the transmit signal; and

halting the adaptive filtering of the low-frequency portion of the transmit signal in response to defecting saturation of the delayed samples of the attenuated receive signal.

28. The method of claim 24, wherein the bandsplitting of the transmit signal and the receive signal each comprises bandsplitting the transmit signal and the receive signal, respectively, into a low-frequency portion having a frequency that is less than or equal to 3400 Hz and a high-frequency portion having a frequency that is between 3400 Hz and 8000 Hz, and wherein the first sample frequency is approximately 16 kHz and the second sample frequency is approximately 8 kHz.

29. The method of claim 24, wherein the bandsplitting of the transmit signal comprises low-pass filtering the transmit signal to generate the low-frequency portion of the transmit signal and subtracting the low-frequency portion of the transmit signal from the transmit signal to generate the high-frequency portion of the transmit signal, and wherein the bandsplitting of the receive signal comprises low-pass filtering the receive signal to generate the low-frequency portion of the receive signal and subtracting the low-frequency portion of the receive signal from the receive signal to generate the high-frequency portion of the receive signal.