Hotword suppression
A method includes adding, by a first computing device, a first audio watermark to first speech data corresponding to playback of a first utterance including a hotword used to invoke an attention of a second computing device. The method includes outputting, by the first computing device, the playback of the first utterance corresponding to the watermarked first speech data. The second computing device is configured to receive the watermarked first speech data and determine to cease processing of the watermarked first speech data.
1 . A computer-implemented method of processing audio data, the method comprising:
receiving, at a computing device, audio data captured by a microphone of the computing device, the audio data comprising an audio watermark that is imperceptible to a human listener;
determining, by an audio watermark identification model executing on the computing device, the audio data includes a predefined audio watermark pattern;
based on determining the audio data includes the predefined audio watermark pattern, determining, by the audio watermark identification model executing on the computing device, the audio data includes the audio watermark; and
based on determining the audio data includes the audio watermark, determining, by the computing device, that processing of the audio data is not to be continued.
2 . The method of claim 1 , wherein the audio watermark identification model comprises a neural network.
3 . The method of claim 2 , wherein the audio watermark pattern includes a replication of a watermark sign matrix, and wherein the neural network is trained to estimate a cross-correlation of the watermark sign matrix.
4 . The method of claim 2 , wherein the neural network is trained using a multi-task loss function.
5 . The method of claim 2 , wherein an output of the neural network is processed by using a filter.
6 . The method of claim 1 , wherein the predefined watermark pattern includes a frequency pattern.
7 . An apparatus, comprising:
processing circuitry configured to
receive audio data captured by a microphone of the computing device, the audio data comprising an audio watermark that is imperceptible to a human listener;
determine, by an audio watermark identification model executing on the processing circuitry, the audio data includes a predefined audio watermark pattern;
based on determining the audio data includes the predefined audio watermark pattern, determine, by the audio watermark identification model, the audio data includes the audio watermark, and
based on determining the audio data includes the audio watermark, determine that processing of the audio data is not to be continued.
8 . The apparatus of claim 7 , wherein the audio watermark identification model comprises a neural network.
9 . The apparatus of claim 8 , wherein the audio watermark pattern includes a replication of a watermark sign matrix, and wherein the neural network is trained to estimate a cross-correlation of the watermark sign matrix.
10 . The apparatus of claim 8 , wherein the neural network is trained using a multi-task loss function.
11 . The apparatus of claim 8 , wherein an output of the neural network is processed by using a filter.
12 . The apparatus of claim 7 , wherein the predefined watermark pattern includes a frequency pattern.
13 . The apparatus of claim 7 , wherein the processing circuitry is configured to:
determine that the audio data includes the audio watermark based on the audio data including the predefined audio watermark pattern; and
determine that the processing of the audio data is to be continued based on the audio data including the audio watermark.
14 . The apparatus of claim 7 , wherein the processing circuitry is configured to:
determine that the audio data does not include the audio watermark based on the audio data not including the predefined audio watermark pattern; and
determine that the processing of the audio data is not to be continued based on the audio data not including the audio watermark.
15 . A non-transitory computer-readable storage medium including computer executable instructions, wherein the instructions, when executed by a computer, cause the computer to perform a method, the method comprising:
receiving, at a computing device, audio data captured by a microphone of the computing device, the audio data comprising an audio watermark that is imperceptible to a human listener;
determining, by an audio watermark identification model executing on the computing device, the audio data includes a predefined audio watermark pattern;
based on determining the audio data includes the predefined audio watermark pattern, determining, by the audio watermark identification model executing on the computing device, the audio data includes the audio watermark; and
based on determining the audio data includes the audio watermark, determining, by the computing device, that processing of the audio data is not to be continued.
16 . The non-transitory computer-readable storage medium of claim 15 , wherein the audio watermark identification model comprises a neural network.
17 . The non-transitory computer-readable storage medium of claim 16 , wherein the audio watermark pattern includes a replication of a watermark sign matrix, and wherein the neural network is trained to estimate a cross-correlation of the watermark sign matrix.
18 . The non-transitory computer-readable storage medium of claim 16 , wherein the neural network is trained using a multi-task loss function.