Determining and adapting to changes in microphone performance of playback devices

1. A method, comprising:

receiving an audio input via a microphone array of a network microphone device, the microphone array comprising a plurality of individual microphones;

producing output microphone signals from each of the individual microphones based on the audio input;

analyzing the output microphone signals to detect a wake word associated with a voice assistant service (VAS);

deriving frequency-domain microphone performance data for each of the individual microphones based on the output microphone signals;

capturing a voice input based on at least one of the output microphone signals; and

after detecting the wake word, (i) sending the voice input to one or more computing devices associated with the VAS, and (ii) sending the frequency-domain microphone performance data, without exposing the audio input from which the microphone performance data is derived, to a remote evaluator to detect aberrant behavior of one or more of the individual microphones, the remote evaluator being distinct from the VAS;

receiving an indication of the aberrant behavior of one or more of the individual microphones; and

modifying a beamforming algorithm of the network microphone device to at least partially compensate for the aberrant behavior.

2. The method of claim 1 , further comprising, at the remote evaluator, comparing the frequency-domain microphone performance data to detect the aberrant behavior.

3. The method of claim 2 , wherein comparing the frequency-domain microphone performance data comprises:

analyzing a frequency response for each of the individual microphones; and

comparing the frequency responses for each of the individual microphones.

4. The method of claim 3 , wherein comparing the frequency responses comprises:

determining a mean or median frequency response for each of the individual microphones; and

identifying any microphone for which the frequency response deviates from the mean or median frequency response by more than a threshold amount.

5. The method of claim 1 , wherein the microphone performance data comprises a windowed average of frequency response data in predefined bands within a frequency range.

6. The method of claim 5 , wherein the frequency range is between about 100 Hz and about 8 kHz.

7. A non-transitory computer-readable medium comprising instructions for identifying aberrant microphone behavior, the instructions, when executed by a processor, causing the processor to perform the following operations:

receiving output microphone signals from individual microphones of a network microphone device;

analyzing the output microphone signals to detect a wake word associated with a voice assistant service (VAS);

deriving frequency-domain microphone performance data for each of the individual microphones based on the output microphone signals;

capturing a voice input based on at least one of the output microphone signals;

after detecting the wake word, (i) sending the voice input to one or more computing devices associated with the VAS, and (ii) sending the frequency-domain microphone performance data, without exposing the output microphone signals from which the microphone performance data is derived, to a remote evaluator to detect aberrant behavior of one or more of the individual microphones, the remote evaluator being distinct from the VAS;

receiving, from the remote evaluator, an indication of the aberrant behavior of one or more of the individual microphones; and

performing corrective action to at least partially compensate for the aberrant behavior.

8. The computer-readable medium of claim 7 , wherein the operations further comprise, after detecting the wake word:

receiving additional audio input via the plurality of individual microphones; and

passing the additional audio input to the one or more computing devices associated with the VAS.

9. The computer-readable medium of claim 7 , wherein the microphone performance data comprises a windowed average of frequency response data in predefined bands within a frequency range.

10. The computer-readable medium of claim 9 , wherein the frequency range is between about 100 Hz and about 8 kHz.

11. A network microphone device comprising:

one or more processors;

a microphone array comprising a plurality of individual microphones; and

a computer-readable medium storing instructions that, when executed by the one or more processors, cause the network microphone device to perform operations, the operations comprising:

receiving output microphone signals from each of the microphones;

analyzing the output microphone signals to detect a wake word associated with a voice assistant service (VAS);

deriving frequency-domain microphone performance data for each of the individual microphones based on the output microphone signals;

capturing a voice input based on at least one of the output microphone signals;

after detecting the wake word, (i) sending the voice input to one or more computing devices associated with the VAS, and (ii) sending the frequency-domain microphone performance data, without exposing the output microphone signals from which the microphone performance data is derived, to a remote evaluator to detect aberrant behavior of one or more of the individual microphones;

receiving an indication of the aberrant behavior of one or more of the individual microphones; and

modifying a beamforming algorithm of the network microphone device to at least partially compensate for the aberrant behavior.

12. The device of claim 11 , wherein the operations further comprise, after detecting the wake word:

receiving additional audio input via the plurality of individual microphones; and

passing the additional audio input to the one or more computing devices associated with the VAS.

13. The device of claim 11 , wherein the microphone performance data comprises a windowed average of frequency response data in predefined bands within a frequency range.

14. The device of claim 13 , wherein the frequency range is between about 100 Hz and about 8 kHz.