Direction based end-pointing for speech recognition
A speech recognition system utilizing automatic speech recognition techniques such as end-pointing techniques in conjunction with beamforming and/or signal processing to isolate speech from one or more speaking users from multiple received audio signals and to detect the beginning and/or end of the speech based at least in part on the isolation. Audio capture devices such as microphones may be arranged in a beamforming array to receive the multiple audio signals. Multiple audio sources including speech may be identified in different beams and processed.
1. A computer-implemented method, comprising:
receiving audio signals from a microphone array, the audio signals representing at least first speech;
processing the audio signals to generate a first beamformed signal;
determining, based on at least one characteristic of the first beamformed signal, that the first beamformed signal represents at least second speech; and
based at least in part on determining that the first beamformed signal represents the second speech, causing speech processing to be performed based at least in part on the first beamformed signal to determine speech processing data corresponding to at least a portion of the second speech.
2. The computer-implemented method of claim 1 , further comprising:
determining a direction corresponding to a source of the audio signals,
wherein processing the audio signals to generate the first beamformed signal comprises processing the audio signals to focus on audio originating from the direction.
3. The computer-implemented method of claim 2 , further comprising:
determining a first energy level of a portion of the audio signals corresponding to the direction; and
determining that the first energy level satisfies a condition.
4. The computer-implemented method of claim 1 , further comprising:
performing wakeword detection to determine the audio signals represent at least the first speech.
5. The computer-implemented method of claim 4 , wherein the causing the speech processing to be performed based at least in part on the first beamformed signal is further based at least in part on the wakeword detection.
6. The computer-implemented method of claim 1 , wherein receiving the audio signals from the microphone array comprises receiving the audio signals from a circular microphone array.
7. The computer-implemented method of claim 1 , wherein receiving the audio signals from the microphone array comprises receiving the audio signals from a linear microphone array.
8. The computer-implemented method of claim 1 , further comprising:
detecting audio using a headset comprising the microphone array.
9. The computer-implemented method of claim 1 , further comprising:
performing echo cancellation to remove noise data from the audio signals to generate the first beamformed signal.
10. The computer-implemented method of claim 1 , further comprising:
determining, based at least in part on the first beamformed signal, an endpoint corresponding to the second speech; and
causing the speech processing to terminate based at least in part on the endpoint.
11. A system, comprising:
a microphone array;
at least one processor; and
at least one memory comprising instructions that, when executed by the at least one processor, cause the system to:
receive audio signals from the microphone array,
process the audio signals to generate a first beamformed signal,
determine, based on at least one characteristic of the first beamformed signal, that the first beamformed signal represents speech,
and
based at least in part on determining that the first beamformed signal represents speech, cause speech processing to be performed based at least in part on the first beamformed signal to determine speech processing data corresponding to at least a portion of the speech.
12. The system of claim 11 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
determine a direction corresponding to a source of the audio signals,
wherein the instructions that cause the system to process the audio signals to generate the first beamformed signal comprise instructions that, when executed by the at least one processor, further cause the system to process the audio signals to focus on audio originating from the direction.
13. The system of claim 12 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
determine a first energy level of a portion of the audio signals corresponding to the direction; and
determine the first energy level satisfies a condition.
14. The system of claim 11 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
perform wakeword detection to determine the audio signals further represent a wakeword.
15. The system of claim 14 , wherein the at least one memory comprises additional instructions that, when executed by the at least one processor, further cause the system to cause the speech processing to be performed further based at least in part on the audio signals further representing the wakeword.
16. The system of claim 11 , wherein the microphone array comprises a circular microphone array.
17. The system of claim 11 , wherein the microphone array comprises a linear microphone array.
18. The system of claim 11 , wherein:
the system further comprises a headset comprising the microphone array; and
the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to detect audio using the headset.
19. The system of claim 11 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
perform echo cancellation to remove noise data from the audio signals to generate the first beamformed signal.
20. The system of claim 11 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
determine, based at least in part on the first beamformed signal, an endpoint corresponding to the speech; and
cause the speech processing to terminate based at least in part on the endpoint.
21. The computer-implemented method of claim 1 , wherein the first speech corresponds to a wakeword associated with the second speech.
22. The computer-implemented method of claim 4 , wherein the first speech corresponds to a wakeword that is detected by the wakeword detection.
23. The computer-implemented method of claim 5 , wherein the first speech corresponds to a wakeword that is detected by the wakeword detection.