Linear transformation for speech recognition modeling
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech recognition using complex linear projection are disclosed. In one aspect, a method includes the actions of receiving audio data corresponding to an utterance. The method further includes generating frequency domain data using the audio data. The method further includes processing the frequency domain data using complex linear projection. The method further includes providing the processed frequency domain data to a neural network trained as an acoustic model. The method further includes generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data.
1. A computer-implemented method comprising:
receiving, by one or more computers, audio data corresponding to an utterance;
generating, by the one or more computers, frequency domain data using the audio data;
processing, by the one or more computers, the frequency domain data using a linear transformation, the linear transformation comprising a complex linear projection;
providing, by the one or more computers, the processed frequency domain data to a neural network of a speech recognition model; and
generating, by the one or more computers, a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data.
2. The method of claim 1 , wherein the speech recognition model comprises an acoustic model.
3. The method of claim 1 , wherein processing the frequency domain data using linear transformation comprises processing the frequency domain data for each of multiple input frames of the audio data.
4. The method of claim 1 , further comprising applying logarithmic compression to the processed frequency domain data before providing the processed frequency domain data to the neural network.
5. The method of claim 1 , wherein the neural network comprises one or more long-short term memory layers and multiple hidden layers.
6. The method of claim 1 , wherein processing the frequency domain data using linear transformation comprises performing a function substantially equivalent to weighted average pooling in the time domain.
7. The method of claim 1 , wherein generating the frequency domain data using the audio comprises applying a Fourier transform to the audio data.
8. The method of claim 1 , wherein generating the frequency domain data using the audio comprises converting the audio data to the frequency domain data.
9. A computer-implemented method comprising:
receiving, by one or more computers, audio data corresponding to an utterance;
generating, by the one or more computers, frequency domain data using the audio data;
processing, by the one or more computers, the frequency domain data using a linear transformation;
providing, by the one or more computers, the processed frequency domain data to a neural network of a speech recognition model;
generating, by the one or more computers, a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data;
generating, by the one or more computers, a convolutional filter with one or more real filter weights; and
generating, by the one or more computers, a frequency domain filter with one or more complex filter weights based on the convolutional filter,
wherein processing the frequency domain data using linear transformation comprises applying the frequency domain filter to the audio data.
10. The method of claim 9 , further comprising:
receiving an additional input frame of audio data corresponding to the utterance;
generating additional frequency domain data using the additional input frame of audio data;
applying the frequency domain filter with the one or more complex filter weights to the additional frequency domain data; and
providing the filtered additional frequency domain data to the neural network,
wherein generating the transcription for the utterance comprises generating the transcription for the utterance based further on output that the neural network provides in response to receiving the filtered additional frequency domain data.
11. The method of claim 9 , wherein the convolutional filter or the frequency domain filter are jointly trained with the neural network.
12. A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
receiving audio data corresponding to an utterance;
generating frequency domain data using the audio data;
processing the frequency domain data using linear transformation, the linear transformation comprising a complex linear projection;
providing the processed frequency domain data to a neural network of a speech recognition model; and
generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data.
13. The system of claim 12 , wherein the speech recognition model comprises an acoustic model.
14. The system of claim 12 , wherein the operations further comprise applying logarithmic compression to the processed frequency domain data before providing the processed frequency domain data to the neural network.
15. The system of claim 12 , wherein the neural network comprises one or more long-short term memory layers and multiple hidden layers.
16. A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
receiving audio data corresponding to an utterance;
generating frequency domain data using the audio data;
processing the frequency domain data using a linear transformation;
providing the processed frequency domain data to a neural network of a speech recognition model;
generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data;
generating a convolutional filter with one or more real filter weights; and
generating a frequency domain filter with one or more complex filter weights based on the convolutional filter,
wherein processing the frequency domain data using linear transformation comprises applying the frequency domain filter to the audio data.
17. The system of claim 16 , wherein the operations further comprise:
receiving an additional input frame of audio data corresponding to the utterance;
generating additional frequency domain data using the additional input frame of audio data;
applying the frequency domain filter with the one or more complex filter weights to the additional frequency domain data; and
providing the filtered additional frequency domain data to the neural network,
wherein generating the transcription for the utterance comprises generating the transcription for the utterance based further on output that the neural network provides in response to receiving the filtered additional frequency domain data.
18. A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
receiving audio data corresponding to an utterance;
generating frequency domain data using the audio data;
processing the frequency domain data using a linear transformation;
providing the processed frequency domain data to a neural network of a speech recognition model; and
generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data,
wherein processing the frequency data comprises applying the linear transformation based on joint training of a filter with the neural network.
19. One or more non-transitory computer-readable media storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:
receiving audio data corresponding to an utterance;
generating frequency domain data using the audio data;
processing the frequency domain data using linear transformation;
providing the processed frequency domain data to a neural network of a speech recognition model;
generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data;
generating a convolutional filter with one or more real filter weights; and
generating a frequency domain filter with one or more complex filter weights based on the convolutional filter,
wherein processing the frequency domain data using linear transformation comprises applying the frequency domain filter to the audio data.