Audio source separation for audio devices

1. A method, comprising:

receiving an audio input at a first device;

transmitting audio information corresponding to the audio input from the first device to a second device;

receiving, responsive to transmitting the audio information, a source separation model from the second device at the first device;

analyzing, at the first device, additional audio input using the received source separation model; and

removing, by the first device, at least a portion of the additional audio input based on the analyzing using the source separation model.

2. The method of claim 1 , further comprising providing an audio output from the first device based on a remaining portion of the additional audio input.

3. The method of claim 1 , wherein the source separation model corresponds to a noise condition identified by the second device in the audio information.

4. The method of claim 1 , wherein the source separation model is one of several source separation models stored at the second device, each source separation model corresponding to a respective noise condition.

5. The method of claim 1 , wherein receiving the source separation model includes receiving a set of weights for a network architecture at the first device.

6. The method of claim 1 , wherein the first device is an earbud and the second device is a companion device that is paired with the earbud.

7. The method of claim 6 , wherein the companion device is a wearable device, a smartphone or a tablet device.

8. The method of claim 1 , further comprising:

receiving further audio input at a third device, wherein the first and third devices are first and second earbuds of a pair of earbuds, and wherein the second device is a companion device that is wirelessly connected to the pair of earbuds;

transmitting further audio information corresponding to the further audio input from the third device to the companion device;

receiving, responsive to providing the further audio information, an additional source separation model from the companion device at the third device;

analyzing, at the third device and concurrently with the analyzing at the first device, further additional audio input using the received additional source separation model; and

removing, by the third device and concurrently with the removing at the first device, at least a portion of the further additional audio input based on the analyzing using the additional source separation model.

9. The method of claim 8 , wherein the source separation model corresponds to a noise condition identified by the companion device in the audio information and the additional source separation model corresponds to a different noise condition identified by the companion device in the further audio information.

10. The method of claim 1 , further comprising:

detecting, with the first device, a change in a noise condition with a classification model at the first device;

providing an indication of the detected change to the second device; and

receiving, from the second device following the indication, a different source separation model.

11. A method, comprising:

receiving audio information at an electronic device from an audio device over a wireless connection;

providing the audio information to a first machine learning model at the electronic device;

obtaining a second machine learning model based on an output of the first machine learning model; and

providing the second machine learning model to the audio device from the electronic device over the wireless connection.

12. The method of claim 11 , further comprising providing a perceptual goal for the audio information as an additional input to the first machine learning model.

13. The method of claim 12 , further comprising:

providing, with the electronic device, a user interface that includes a tuner having controls that are operable by a user of the electronic device to control enhancement or suppression of one or more sound types;

receiving an input to the tuner from the user;

modifying the perceptual goal according to the input to the tuner;

obtaining a new output of the first machine learning model based at least in part on the modified perceptual goal; and

providing a third machine learning model to the audio device based on the new output of the first machine learning model.

14. The method of claim 12 , further comprising determining the perceptual goal based at least in part on a context of the electronic device.

15. The method of claim 11 , wherein the output of the first machine learning model comprises an identifier of one of a plurality of machine learning models that are stored at the electronic device.

16. The method of claim 11 , wherein the output of the first machine learning model comprises an identifier of a noise condition associated with the audio information, and wherein obtaining the second machine learning model includes obtaining a pre-stored machine learning model from a plurality of pre-stored machine learning models at the electronic device using the identifier of the noise condition.

17. The method of claim 11 , wherein the output of the first machine learning model comprises a set of parameters for a network architecture stored at the audio device, and wherein providing the second machine learning model to the audio device comprises providing the set of parameters to the audio device.

18. The method of claim 11 , further comprising:

receiving additional audio information at the electronic device from another audio device;

providing the additional audio information to the first machine learning model at the electronic device;

obtaining a third machine learning model based on an additional output of the first machine learning model using the additional audio information; and

providing the third machine learning model to the additional audio device from the electronic device.

19. The method of claim 18 , wherein the second machine learning model corresponds to a noise condition identified by the first machine learning model in the audio information, and wherein the third machine learning model corresponds to a different noise condition identified by the first machine learning model in the additional audio information.

20. The method of claim 18 , wherein the second machine learning model and the third machine learning model are the same machine learning model and both correspond to a noise condition identified by the first machine learning model in the audio information and the additional audio information.

21. The method of claim 18 , wherein the audio device and the additional audio device are first and second earbuds of a pair of earbuds, and wherein the electronic device is a companion device that is wirelessly connected to the pair of earbuds over the wireless connection.

22. The method of claim 11 , further comprising:

detecting a change in a noise condition of the electronic device at least in part using a sensor of the electronic device;

requesting additional audio information from the audio device responsive to detecting the change;

providing the additional audio information to the first machine learning model at the electronic device;

obtaining a third machine learning model based on an additional output of the first machine learning model; and

providing the third machine learning model to the audio device from the electronic device over the wireless connection.

23. An audio device, comprising:

at least one microphone; and

a processor, wherein the processor is configured to:

receive audio input using the at least one microphone;

transmit audio information corresponding to the audio input to a companion device;

receive, responsive to transmitting the audio information, a source separation model from the companion device;

analyze additional audio input received by the microphone using the received source separation model; and

remove at least a portion of the additional audio input based on the analysis using the received source separation model.