System and method for audio-visual multi-speaker speech separation with location-based selection

1. A method for audio-visual multi-speaker speech separation, comprising:

receiving audio signals captured by at least one microphone;

receiving video signals captured by at least one camera; and

providing the audio signals and the video signals to a sync engine configured to:

derive an audio vector from the audio signals and a video vector from the video signals;

compute a correlation score by shifting either the audio vector or the video vector and compare the shifted vector against a remaining unshifted vector, wherein the correlation score is based on a number of shifts needed to achieve a match; and

extract facial characteristics of each speaker from multi-speaker synchronized video signals to provide for mutual influence between audio and video to assist in an audio-visual separation;

apply audio-visual separation on the received audio signals and the video signals by simultaneously analyzing each of multi-speaker synchronized video signals to provide isolation of sounds from the at least one microphone and the at least one camera based on the correlation score by generating an audio output comprising any of:

a time-shifted variant of the audio signal based on a number of shifts of the audio signal assigned a highest correlation score;

a time-shifted variant of the video signal based on a number of shifts of the video signal assigned a highest correlation score; and

the audio signal time-shifted to synchronize with lip movements in the video signal,

wherein the audio-visual separation is based, in part, on angle positions of at least one speaker relative to the at least one camera.

2. The method of claim 1 , further comprising:

determining, based on gesture recognition, the intention of a speaker to talk.

3. The method of claim 1 , further comprising:

separating the audio signals into multiple distinct voice and noise channels by their contents.

4. The method of claim 1 , further comprising performing at least one of:

applying echo cancellation on the received audio signals; or

applying synchronization correction on the received audio signals.

5. The method of claim 1 , wherein the at least one microphone includes an array of microphones, and wherein the array of microphones is directed to a specific position in the space based on the angle positions.

6. The method of claim 5 , further comprising:

generating a beamformer control signal to control the aiming direction of the array of microphones in the space based on a facial recognition.

7. The method of claim 5 , further comprising:

applying speech recognition on each separated audio channel.

8. The method of claim 7 , further comprising:

applying natural language processing on recognized speech to extract user intention.

9. The method of claim 8 , further comprising:

personalizing voice commands based on the separated voice channels.

10. The method of claim 1 , wherein applying audio-visual separation further comprises:

detecting faces appearing in the captured video signals;

tracking each detected face; and

determining angle positions based on each tracked face, wherein the current speaker is determined by a tracked face.

11. The method of claim 10 , wherein determining angle positions includes analysis of information on at least: image attributes captured by the at least one camera, intrinsic parameters of the at least one camera, and factors describing the angle and position of the sound source.

12. The method of claim 11 , wherein the image attributes further include:

image coordinates, wherein the image coordinates define at least a position of a set of lips in the captured video, and undistorted image coordinates, wherein the undistorted image coordinates define a position of a set of lips in an undistorted image.

13. The method of claim 11 , further comprising:

converting the image coordinates into world coordinates; and

determining a camera mouth angle position in a tracked face relative to the at least one camera based on the world coordinates.

14. The method of claim 11 , further comprising:

determining a microphone camera angle position of the at least one camera relative to the at least one microphone; and

determining a mouth angle position in a tracked face relative to the at least one microphone based on the camera mouth angle position and the microphone camera angle position.

15. A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to execute a process for audio-visual multi-speaker speech separation, the process comprising:

receiving audio signals captured by at least one microphone;

receiving video signals captured by at least one camera; and

providing the audio signals and the video signals to a sync engine configured to:

derive an audio vector from the audio signals and the video vector from the video signals;

compute a correlation score by shifting either the audio vector or a video vector and compare the shifted vector against a remaining unshifted vector, wherein the correlation score is based on a number of shifts needed to achieve a match; and

extract facial characteristics of each speaker from multi-speaker synchronized video signals provide for mutual influence between audio and video to assist in an audio-visual separation;

apply audio-visual separation on the received audio signals and the video signals by simultaneously analyzing each of multi-speaker synchronized video signals to provide isolation of sounds from the at least one microphone and the at least one camera based on the correlation score by generating an audio output comprising any of:

a time-shifted variant of the audio signal based on a number of shifts of the audio signal assigned a highest correlation score;

a time-shifted variant of the video signal based on a number of shifts of the video signal assigned a highest correlation score; and

the audio signal time-shifted to synchronize with lip movements in the video signal,

wherein the audio-visual separation is based, in part, on angle positions of at least one speaker relative to the at least one camera.

16. A system for audio-visual multi-speaker speech separation, comprising:

a processing circuitry; and

a memory, the memory containing instructions that, when executed by the processing circuitry, configure the system to:

receive audio signals captured by at least one microphone;

receive video signals captured by at least one camera; and

provide the audio signals and the video signals to a sync engine configured to:

derive an audio vector from the audio signals and a video vector from the video signals;

compute a correlation score by shifting either the audio vector or the video vector and compare the shifted vector against a remaining unshifted vector, wherein the correlation score is based on a number of shifts needed to achieve a match; and

extract facial characteristics of each speaker from multi-speaker synchronized video signals to provide for mutual influence between audio and video to assist in an audio-visual separation;

apply audio-visual separation on the received audio signals and the video signals by simultaneously analyzing each of multi-speaker synchronized video signals to provide isolation of sounds from the at least one microphone and the at least one camera based on the correlation score by generating an audio output comprising any of:

a time-shifted variant of the audio signal based on a number of shifts of the audio signal assigned a highest correlation score;

a time-shifted variant of the video signal based on a number of shifts of the video signal assigned a highest correlation score; and

the audio signal time-shifted to synchronize with lip movements in the video signal,

wherein the audio-visual separation is based, in part, on angle positions of at least one speaker relative to the at least one camera.

17. The system of claim 16 , wherein the system is further configured to:

determine, based on gesture recognition, the intention of a speaker to talk.

18. The system of claim 16 , wherein the system is further configured to:

separate the audio signals into multiple distinct voice and noise channels by their contents.

19. The system of claim 16 , wherein the system is further configured to apply at least one of:

echo cancellation on the received audio signals; or

synchronization correction on the received audio signals.

20. The system of claim 16 , wherein the at least one microphone includes an array of microphones, and wherein the array of microphones is directed to a specific position in the space based on the angle positions.

21. The system of claim 20 , wherein the system is further configured to:

generate a beamformer control signal to control the aiming direction of the array of microphones in the space based on a facial recognition.

22. The system of claim 20 , wherein the system is further configured to:

apply speech recognition on each separated audio channel.

23. The system of claim 22 , wherein the system is further configured to:

apply natural language processing on recognized speech to extract user intention.

24. The system of claim 23 , wherein the system is further configured to:

personalize voice commands based on the separated voice channels.

25. The system of claim 16 , wherein the system is further configured to:

detect faces appearing in the received video signals;

track each detected face; and

determine angle positions based on each tracked face, wherein the current speaker is determined by a tracked face.

26. The system of claim 25 , wherein determining angle positions includes analysis of information on at least: image attributes captured by the at least one camera, intrinsic parameters of the at least one camera, and factors describing the angle and position of the sound source.

27. The system of claim 26 , wherein the image attributes further include:

image coordinates, wherein the image coordinates define at least a position of a set of lips in the captured video, and undistorted image coordinates, wherein the undistorted image coordinates define a position of a set of lips in an undistorted image.

28. The system of claim 26 , wherein the system is further configured to:

convert the image coordinates into world coordinates; and

determine a camera mouth angle position in a tracked face relative to the at least one camera based on the world coordinates.

29. The system of claim 26 , wherein the system is further configured to:

determine a microphone camera angle position of the at least one camera relative to the at least one microphone; and

determine a mouth angle position in a tracked face relative to the at least one microphone based on the camera mouth angle position and the microphone camera angle position.