Cascade echo cancellation for asymmetric references
A system configured to perform cascade echo cancellation processing to improve a performance when reference signals are asymmetric (e.g., dominant reference signal(s) overshadow weak reference signal(s)). The system may perform cascade echo cancellation processing to separately adapt filter coefficients between the dominant reference signal(s) and the weak reference signal(s). For example, the system may use a dominant reference signal to process a microphone audio signal and generate a residual audio signal, using the residual audio signal to adapt first filter coefficient values corresponding to the dominant reference signal. Separately, the system may use a weak reference signal to process the residual audio signal and generate an output audio signal, using the output audio signal to adapt second filter coefficient values corresponding to the weak reference signal.
1. A computer-implemented method, the method comprising:
receiving first reference audio data associated with a first loudspeaker corresponding to a device;
receiving second reference audio data associated with a second loudspeaker corresponding to the device;
receiving first audio data from at least one microphone of the device;
determining that the first reference audio data represents a stronger signal than the second reference audio data;
based at least in part on determining that the first reference audio data represents a stronger signal than the second reference audio data, generating second audio data by performing first echo cancellation using the first audio data and the first reference audio data; and
based at least in part on determining that the first reference audio data represents a stronger signal than the second reference audio data, generating third audio data by performing second echo cancellation using the second audio data and the second reference audio data.
2. The computer-implemented method of claim 1 , further comprising:
determining a first energy value of a portion of the first reference audio data within a first frequency range; and
determining a second energy value of a portion of the second reference audio data within the first frequency range,
wherein determining that the first reference audio data represents a stronger signal than the second reference audio data comprises determining that a first difference between the first energy value and the second energy value satisfies a condition.
3. The computer-implemented method of claim 2 , further comprising:
receiving fourth audio data from the at least one microphone;
determining a third energy value of a second portion of the first reference audio data within a second frequency range;
determining a fourth energy value of a second portion of the second reference audio data within the second frequency range;
determining that a second difference between the third energy value and the fourth energy value does not satisfy the condition; and
generating fifth audio data using the fourth audio data, the second portion of the first reference audio data and the second portion of the second reference audio data.
4. The computer-implemented method of claim 1 , wherein:
the first echo cancellation is performed using a first plurality of filter coefficient values associated with a first echo canceller; and
the second echo cancellation is performed using a second plurality of filter coefficient values associated with a second echo canceller.
5. The computer-implemented method of claim 4 , further comprising:
generating a third plurality of filter coefficient values using the second audio data and the first plurality of filter coefficient values.
6. The computer-implemented method of claim 4 , further comprising:
generating a third plurality of filter coefficient values using the third audio data and the second plurality of filter coefficient values.
7. The computer-implemented method of claim 1 , further comprising:
receiving third reference audio data associated with the first loudspeaker;
receiving fourth reference audio data associated with the second loudspeaker;
receiving fourth audio data from the at least one microphone;
determining that a difference between the third reference audio data and the fourth reference audio data fails to satisfy a condition; and
generating fifth audio data by performing third echo cancellation using the fourth audio data, the third reference audio data, and the fourth reference audio data.
8. The computer-implemented method of claim 1 , further comprising:
receiving third reference audio data associated with a third loudspeaker of the device; and
determining that a difference between the first reference audio data and the third reference audio data fails to satisfy a condition,
wherein performing the first echo cancellation comprises:
generating first estimated echo audio data using the first reference audio data and a first portion of a first plurality of filter coefficient values,
generating second estimated echo audio data using the third reference audio data and a second portion of the first plurality of filter coefficient values, and
generating the second audio data using the first audio data, the first estimated echo audio data, and the second estimated echo audio data.
9. The computer-implemented method of claim 8 , wherein generating the second audio data comprises:
generating the second audio data by subtracting the first estimated echo audio data and the second estimated echo audio data from the first audio data.
10. The computer-implemented method of claim 1 , further comprising:
receiving third reference audio data associated with the first loudspeaker;
receiving fourth reference audio data associated with the second loudspeaker;
receiving fourth audio data from the at least one microphone;
determining a first signal quality metric value of the third reference audio data;
determining a second signal quality metric value of the fourth reference audio data;
determining that the first signal quality metric value satisfies a condition;
determining that the second signal quality metric value fails to satisfy the condition; and
generating fifth audio data by performing third echo cancellation using the fourth audio data and the fourth reference audio data.
11. A system comprising:
at least one processor; and
memory including instructions operable to be executed by the at least one processor to cause the system to:
receive first reference audio data associated with a first loudspeaker corresponding to a device;
receive second reference audio data associated with a second loudspeaker corresponding to the device;
receive first audio data from at least one microphone of the device;
determine that the first reference audio data represents a stronger signal than the second reference audio data;
based at least in part on determining that the first reference audio data represents a stronger signal than the second reference audio data, generate second audio data by performing first echo cancellation using the first audio data and the first reference audio data; and
based at least in part on determining that the first reference audio data represents a stronger signal than the second reference audio data, generate third audio data by performing second echo cancellation using the second audio data and the second reference audio data.
12. The system of claim 11 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
determine a first energy value of a portion of the first reference audio data within a first frequency range; and
determine a second energy value of a portion of the second reference audio data within the first frequency range,
wherein the instructions that cause the system to determine that the first reference audio data represents a stronger signal than the second reference audio data comprise instructions that, when executed by the at least one processor, further cause the system to determine that a first difference between the first energy value and the second energy value satisfies a condition.
13. The system of claim 12 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
receive fourth audio data from the at least one microphone;
determine a third energy value of a second portion of the first reference audio data within a second frequency range;
determine a fourth energy value of a second portion of the second reference audio data within the second frequency range;
determine that a second difference between the third energy value and the fourth energy value does not satisfy the condition; and
generate fifth audio data using the fourth audio data, the second portion of the first reference audio data and the second portion of the second reference audio data.
14. The system of claim 11 , wherein:
the first echo cancellation is performed using a first plurality of filter coefficient values associated with a first echo canceller; and
the second echo cancellation is performed using a second plurality of filter coefficient values associated with a second echo canceller.
15. The system of claim 14 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
generate a third plurality of filter coefficient values using the second audio data and the first plurality of filter coefficient values.
16. The system of claim 14 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
generate a third plurality of filter coefficient values using the third audio data and the second plurality of filter coefficient values.
17. The system of claim 11 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
receive third reference audio data associated with the first loudspeaker;
receive fourth reference audio data associated with the second loudspeaker;
receive fourth audio data from the at least one microphone;
determine that a difference between the third reference audio data and the fourth reference audio data fails to satisfy a condition; and
generate fifth audio data by performing third echo cancellation using the fourth audio data, the third reference audio data, and the fourth reference audio data.
18. The system of claim 11 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
receive third reference audio data associated with a third loudspeaker of the device; and
determine that a difference between the first reference audio data and the third reference audio data fails to satisfy a condition,
wherein the instructions that cause the system to perform the first echo cancellation comprise instructions that, when executed by the at least one processor, further cause the system to:
generate first estimated echo audio data using the first reference audio data and a first portion of a first plurality of filter coefficient values,
generate second estimated echo audio data using the third reference audio data and a second portion of the first plurality of filter coefficient values, and
generate the second audio data using the first audio data, the first estimated echo audio data, and the second estimated echo audio data.
19. The system of claim 18 , wherein the instructions that cause the system to generate the second audio data comprise instructions that, when executed by the at least one processor, further cause the system to:
generate the second audio data by subtracting the first estimated echo audio data and the second estimated echo audio data from the first audio data.
20. The system of claim 11 , wherein the memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
receive third reference audio data associated with the first loudspeaker;
receive fourth reference audio data associated with the second loudspeaker;
receive fourth audio data from the at least one microphone;
determine a first signal quality metric value of the third reference audio data;
determine a second signal quality metric value of the fourth reference audio data;
determine that the first signal quality metric value satisfies a condition;
determine that the second signal quality metric value fails to satisfy the condition; and
generate fifth audio data by performing third echo cancellation using the fourth audio data and the fourth reference audio data.