Audio source localization

1. A method comprising:

at an electronic device that includes a microphone array including at least a first microphone and a second microphone:

receiving, via the first microphone, a first signal representative of audio emitted by an audio source over a time period;

receiving, via the second microphone, a second signal representative of the audio emitted by the audio source over the time period;

estimating a first amplitude associated with the first signal and the second signal for a first time within the time period;

estimating a second amplitude associated with the first signal and the second signal for a second time within the time period, the second time being different than the first time;

normalizing the first signal and the second signal at the first time by the first amplitude to determine a first normalized amplitude;

generating a first vector that include a first set of elements, wherein the first vector includes the first normalized amplitude associated with the first time;

normalizing the first signal and the second signal at the second time by the second amplitude to determine a second normalized amplitude;

generating a second vector that include a second set of elements, wherein the second vector includes the second normalized amplitude;

determining a combined amplitude representative of the audio emitted by the audio source by combining the first normalized amplitude and the second normalized amplitude, wherein the combined amplitude includes a third vector comprising the first set of elements and the second set of elements; and

estimating, based at least in part on the combined amplitude, a position of the audio source relative to the device.

2. The method of claim 1 , wherein the time period is a first time period, the method further comprising:

receiving, from the first microphone, a third signal representative of the audio emitted by the audio source over a second time period;

receiving, from the second microphone, a fourth signal representative of the audio emitted by the audio source over the second time period;

determining a change in pose of the device between the first time period and the second time period; and

determining, based at least in part on the third signal, the fourth signal, the change in pose, and the position, an updated position of the audio source relative to the device associated with the second time period.

3. The method of claim 1 , further comprising:

forming multiple beams representative of audio signals in an environment surrounding the device, the audio signals including the first signal and the second signal;

determining, based at least in part on the multiple beams, multiple directions from which the audio signals originate relative to the microphone array;

determining that a direction of the multiple directions has a higher response value than other directions of the multiple directions; and

determining that the first signal and the second signal are associated with the direction,

wherein normalizing the first signal and the second signal at the first time comprises normalizing the first signal and the second signal by an amplitude of the first signal and the second signal associated with the direction, and normalizing the first signal and the second signal at the second time comprises normalizing the first signal and the second signal by the amplitude of the first signal and the second signal associated with the direction.

4. The method of claim 1 , further comprising:

transforming the first signal and the second signal to a frequency domain using a short-time Fourier transform (STFT),

wherein estimating the first amplitude is based at least in part on determining a frequency associated with the first signal and the second signal in the frequency domain at the first time, and

wherein estimating the second amplitude is based at least in part on determining the frequency associated with the first signal and the second signal in the frequency domain at the second time.

5. The method of claim 4 , further comprising:

separating the first signal and the second signal into a first portion of the first signal and the second signal associated with the first time and a second portion of the first signal and the second signal associated with the second time,

wherein transforming the first signal and the second signal to the frequency domain comprises transforming the first portion and the second portion separately using the STFT.

6. The method of claim 1 , further comprising:

generating a spatial spectrum based at least in part on the first signal and the second signal, the spatial spectrum indicating an amount of energy arriving at the microphone array from different directions;

determining a maximum value in the spatial spectrum; and

determining an instantaneous direction of arrival of the audio based at least in part on the maximum value in the spatial spectrum,

wherein estimating the first amplitude and estimating the second amplitude are further based on the instantaneous direction of arrival.

7. The method of claim 1 , further comprising:

determining a localization performance of multiple microphones corresponding to determining the position of the audio source relative to the device;

comparing the localization performance of the multiple microphones to a localization performance threshold; and

responsive to determining that the localization performance of the multiple microphones is greater than the localization performance threshold, using less than all of the multiple microphones to receive signals representative of audio emitted by the audio source.

8. The method of claim 1 , wherein the audio source corresponds to human speech having a varying amplitude over the time period.

9. One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

receiving, from a first microphone of a microphone array of a device, a first signal representative of audio emitted by an audio source;

receiving, from a second microphone of the microphone array of the device, a second signal representative of the audio emitted by the audio source;

estimating a first amplitude associated with the first signal and the second signal for a first time;

estimating a second amplitude associated with the first signal and the second signal for a second time, the second time being different than the first time;

normalizing the first signal and the second signal at the first time by the first amplitude to determine a first normalized amplitude;

generating a first vector that include a first set of elements, wherein the first vector includes the first normalized amplitude associated with the first time;

normalizing the first signal and the second signal at the second time by the second amplitude to determine a second normalized amplitude;

generating a second vector that include a second set of elements, wherein the second vector includes the second normalized amplitude associated with the second time;

determining a combined amplitude by combining the first normalized amplitude and the second normalized amplitude, wherein the combined amplitude includes a third vector comprising the first set of elements and the second set of elements; and

estimating, based at least in part on the combined amplitude, a position of the audio source relative to the device.

10. The one or more non-transitory computer-readable media of claim 9 , the operations further comprising:

transforming the first signal and the second signal to a frequency domain using a short-time Fourier transform (STFT),

wherein estimating the first amplitude is based at least in part on determining a frequency associated with the first signal and the second signal in the frequency domain at the first time, and

wherein estimating the second amplitude is based at least in part on determining the frequency associated with the first signal and the second signal in the frequency domain at the second time.

11. The one or more non-transitory computer-readable media of claim 10 , the operations further comprising:

separating the first signal and the second signal into a first portion of the first signal and the second signal associated with the first time and a second portion of the first signal and the second signal associated with the second time,

wherein transforming the first signal and the second signal to the frequency domain comprises transforming the first portion and the second portion separately using the STFT.

12. The one or more non-transitory computer-readable media of claim 9 , the operations further comprising:

generating a spatial spectrum based at least in part on the first signal and the second signal, the spatial spectrum indicating an amount of energy arriving at the microphone array from different directions;

determining a maximum value in the spatial spectrum; and

determining an instantaneous direction of arrival of the audio based at least in part on the maximum value in the spatial spectrum,

wherein estimating the first amplitude and estimating the second amplitude are further based on the instantaneous direction of arrival.

13. An electronic device comprising:

a microphone array comprising a first microphone and a second microphone;

one or more processors; and

one or more computer-readable media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

receiving, from the first microphone, a first signal representative of audio emitted by an audio source;

receiving, from the second microphone, a second signal representative of the audio emitted by the audio source;

estimating a first representative amplitude of the audio based at least in part on the first signal and the second signal associated with a first time;

estimating a second representative amplitude of the audio based at least in part on the first signal and the second signal associated with a second time, the second time being different than the first time;

normalizing the first signal and the second signal by the first representative amplitude to determine a first normalized amplitude;

generating a first vector that include a first set of elements, wherein the first vector includes the first normalized amplitude associated with the first time;

normalizing the first signal and the second signal by the second representative amplitude to determine a second normalized amplitude;

generating a second vector that include a second set of elements, wherein the second vector includes the second normalized amplitude associated with the second time;

determining a combined amplitude by combining the first normalized amplitude and the second normalized amplitude, wherein the combined amplitude includes a third vector comprising the first set of elements and the second set of elements; and

estimating, based at least in part on the combined amplitude, a position of the audio source relative to the electronic device that includes the microphone array.

14. The electronic device of claim 13 , wherein the first time and the second time are within a first time period, the operations further comprising:

receiving, from the first microphone, a third signal representative of audio emitted by the audio source over a second time period;

receiving, from the second microphone, a fourth signal representative of the audio emitted by the audio source over the second time period;

determining a change in pose of the electronic device between the first time period and the second time period; and

determining, based at least in part on the third signal, the fourth signal, the change in pose of the electronic device, and the position, an updated position of the audio source relative to the electronic device associated with the second time period.

15. The electronic device of claim 13 , the operations further comprising:

forming multiple beams representative of audio signals in an environment surrounding the device, the audio signals including the first signal and the second signal;

determining, based at least in part on the multiple beams, multiple directions from which the audio signals originate relative to the microphone array;

determining that a direction of the multiple directions has a higher response value than other directions of the multiple directions; and

determining that the first signal and the second signal are associated with the direction,

wherein normalizing the first signal and the second signal at the first time comprises normalizing the first signal and the second signal by an amplitude of the first signal and the second signal associated with the direction, and normalizing the first signal and the second signal at the second time comprises normalizing the first signal and the second signal by the amplitude of the first signal and the second signal associated with the direction.

16. The electronic device of claim 13 , wherein the microphone array comprises multiple microphones including the first microphone and the second microphone, the operations further comprising:

determining a localization performance of the multiple microphones corresponding to determining the position of the audio source relative to the electronic device;

comparing the localization performance of the multiple microphones to a localization performance threshold; and

responsive to determining that the localization performance of the multiple microphones is greater than the localization performance threshold, using less than all of the multiple microphones to receive signals representative of audio emitted by the audio source.

17. The electronic device of claim 13 , wherein the audio source corresponds to human speech having a varying amplitude.

18. The electronic device of claim 13 , the operations further comprising:

transforming the first signal and the second signal to a frequency domain using a short-time Fourier transform (STFT),

wherein estimating the first amplitude is based at least in part on determining a frequency associated with the first signal and the second signal in the frequency domain at the first time, and

wherein estimating the second amplitude is based at least in part on determining the frequency associated with the first signal and the second signal in the frequency domain at the second time.

19. The electronic device of claim 13 , the operations further comprising:

separating the first signal and the second signal into a first portion of the first signal and the second signal associated with the first time and a second portion of the first signal and the second signal associated with the second time,

wherein transforming the first signal and the second signal to the frequency domain comprises transforming the first portion and the second portion separately using the STFT.

20. The electronic device of claim 13 , the operations further comprising:

generating a spatial spectrum based at least in part on the first signal and the second signal, the spatial spectrum indicating an amount of energy arriving at the microphone array from different directions;

determining a maximum value in the spatial spectrum; and

determining an instantaneous direction of arrival of the audio based at least in part on the maximum value in the spatial spectrum,

wherein estimating the first amplitude and estimating the second amplitude are further based on the instantaneous direction of arrival.