Query endpointing based on lip detection
Systems and methods are described for improving endpoint detection of a voice query submitted by a user. In some implementations, a synchronized video data and audio data is received. A sequence of frames of the video data that includes images corresponding to lip movement on a face is determined. The audio data is endpointed based on first audio data that corresponds to a first frame of the sequence of frames and second audio data that corresponds to a last frame of the sequence of frames. A transcription of the endpointed audio data is generated by an automated speech recognizer. The generated transcription is then provided for output.
1. A computer-implemented method comprising:
receiving, by one or more computing devices, video data and audio data that is synchronized with the video data;
determining, by one or more computing devices, that a sequence of video frames of the video data includes a representation of a face with moving lips;
determining, by the one or more computing devices and based on analyzing the video data, that the moving lips in the sequence of video frames are associated with an activity other than speech; and
based on determining that the moving lips in the sequence of video frames are associated with an activity other than speech, bypassing, by the one or more computing devices, performing speech recognition on audio data that is synchronized with the sequence of video frames.
2. The method of claim 1 , wherein:
the video data and the audio data that is synchronized with the video data are received from a smartphone; and
the video data is captured by a front-facing camera of the smartphone.
3. The method of claim 1 , comprising:
determining that the video data includes user motion; and
in response to determining that the video data includes user motion, determining that the sequence of video frames includes the representation of the face with moving lips.
4. The method of claim 1 , wherein determining that the sequence of video frames of the video data includes a representation of a face with moving lips comprises:
identifying one or more feature statistics for the representation of the face with moving lips; and
providing the sequence of video frames as input to a model that is trained to determine whether the one or more feature statistics identified for the representation of the face with moving lips include a feature statistic for lip movement associated with speech.
5. The method of claim 1 , wherein determining that the sequence of video frames includes the representation of a face with moving lips comprises:
obtaining multiple subsets of consecutive video frames of the video data; and
processing each of the subsets of consecutive video frames of the video data using a deep neural network configured to:
receive each subset of consecutive video frames;
compute, for each subsets of consecutive video frames, a confidence score that represents a likelihood that the subset of consecutive video frames includes a representation of the face with moving lips; and
determine that each confidence score satisfies a confidence score threshold.
6. The method of claim 1 , wherein determining that the moving lips in the sequence of video frames are associated with the activity other than speech comprises:
obtaining multiple subsets of consecutive video frames of the sequence of video frames;
processing each of the subsets of consecutive video frames of the sequence of video frames using a deep neural network configured to:
receive each subset of consecutive video frames of the sequence of video frames;
compute, for each subset of consecutive video frames of the sequence of video frames, a confidence score that represents a likelihood that the subset of consecutive video frames includes a representation of the face with lip movement associated with speech; and
determine that each confidence score does not satisfy a confidence threshold.
7. The method of claim 1 , wherein bypassing performing speech recognition on audio data that is synchronized with the sequence of video frames comprises preventing generating a transcription of the audio data by an automated speech recognizer.
8. The method of claim 1 , wherein:
the audio data comprises a speech portion corresponding to a voice query provided by a user associated with the face; and
determining that the moving lips in the sequence of video frames are associated with an activity other than speech comprises:
determining that the audio data that is synchronized with the sequence of video frames is not included in the speech portion corresponding to the voice query provided by the user associated with the face.
9. A system comprising:
one or more computing devices; and
one or more storage devices storing instructions that are operable, when executed by the one or more computing devices, to cause the one or more computing devices to perform operations comprising:
receiving, by one or more computing devices, video data and audio data that is synchronized with the video data;
determining, by one or more computing devices, that a sequence of video frames of the video data includes a representation of a face with moving lips;
determining, by the one or more computing devices and based on analyzing the video data, that the moving lips in the sequence of video frames are associated with an activity other than speech; and
based on determining that the moving lips in the sequence of video frames are associated with an activity other than speech, bypassing, by the one or more computing devices, performing speech recognition on audio data that is synchronized with the sequence of video frames.
10. The system of claim 9 , wherein:
the video data and the audio data that is synchronized with the video data are received from a smartphone; and
the video data is captured by a front-facing camera of the smartphone.
11. The system of claim 9 , wherein the operations further comprise:
determining that the video data includes user motion; and
in response to determining that the video data includes user motion, determining that the sequence of video frames includes the representation of the face with moving lips.
12. The system of claim 9 , wherein determining that the sequence of video frames of the video data includes a representation of a face with moving lips comprises:
identifying one or more feature statistics for the representation of the face with moving lips; and
providing the sequence of video frames as input to a model that is trained to determine whether the one or more feature statistics identified for the representation of the face with moving lips include a feature statistic for lip movement associated with speech.
13. The system of claim 9 , wherein determining that the sequence of video frames includes the representation of a face with moving lips comprises:
obtaining multiple subsets of consecutive video frames of the video data; and
processing each of the subsets of consecutive video frames of the video data using a deep neural network configured to:
receive each subset of consecutive video frames;
compute, for each subsets of consecutive video frames, a confidence score that represents a likelihood that the subset of consecutive video frames includes a representation of the face with moving lips; and
determine that each confidence score satisfies a confidence score threshold.
14. The system of claim 9 , wherein determining that the moving lips in the sequence of video frames are associated with the activity other than speech comprises:
obtaining multiple subsets of consecutive video frames of the sequence of video frames;
processing each of the subsets of consecutive video frames of the sequence of video frames using a deep neural network configured to:
receive each subset of consecutive video frames of the sequence of video frames;
compute, for each subset of consecutive video frames of the sequence of video frames, a confidence score that represents a likelihood that the subset of consecutive video frames includes a representation of the face with lip movement associated with speech; and
determine that each confidence score does not satisfy a confidence threshold.
15. A non-transitory computer-readable storage device encoded with computer program instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:
receiving, by one or more computing devices, video data and audio data that is synchronized with the video data;
determining, by one or more computing devices, that a sequence of video frames of the video data includes a representation of a face with moving lips;
determining, by the one or more computing devices and based on analyzing the video data, that the moving lips in the sequence of video frames are associated with an activity other than speech; and
based on determining that the moving lips in the sequence of video frames are associated with an activity other than speech, bypassing, by the one or more computing devices, performing speech recognition on audio data that is synchronized with the sequence of video frames.
16. The non-transitory computer-readable storage device of claim 15 , wherein:
the video data and the audio data that is synchronized with the video data are received from a smartphone; and
the video data is captured by a front-facing camera of the smartphone.
17. The non-transitory computer-readable storage device of claim 15 , wherein the operations further comprise:
determining that the video data includes user motion; and
in response to determining that the video data includes user motion, determining that the sequence of video frames includes the representation of the face with moving lips.
18. The non-transitory computer-readable storage device of claim 15 , wherein determining that the sequence of video frames of the video data includes a representation of a face with moving lips comprises:
identifying one or more feature statistics for the representation of the face with moving lips; and
providing the sequence of video frames as input to a model that is trained to determine whether the one or more feature statistics identified for the representation of the face with moving lips include a feature statistic for lip movement associated with speech.
19. The non-transitory computer-readable storage device of claim 15 , wherein determining that the sequence of video frames includes the representation of a face with moving lips comprises:
obtaining multiple subsets of consecutive video frames of the video data; and
processing each of the subsets of consecutive video frames of the video data using a deep neural network configured to:
receive each subset of consecutive video frames;
compute, for each subsets of consecutive video frames, a confidence score that represents a likelihood that the subset of consecutive video frames includes a representation of the face with moving lips; and
determine that each confidence score satisfies a confidence score threshold.
20. The non-transitory computer-readable storage device of claim 15 , wherein bypassing performing speech recognition on audio data that is synchronized with the sequence of video frames comprises preventing generating a transcription of the audio data by an automated speech recognizer.