Speech recognition power management

Patent metadata

Field	Value
Publication number	US-9704486-B2
Application number	US-201213711510-A
Country	US
Kind code	B2
Filing date	Dec 11, 2012
Priority date	Dec 11, 2012
Publication date	Jul 11, 2017
Grant date	Jul 11, 2017

Abstract

Power consumption for a computing device may be managed by one or more keywords. For example, if an audio input obtained by the computing device includes a keyword, a network interface module and/or an application processing module of the computing device may be activated. The audio input may then be transmitted via the network interface module to a remote computing device, such as a speech recognition server. Alternately, the computing device may be provided with a speech recognition engine configured to process the audio input for on-device speech recognition.

First claim

What is claimed is:

1. A system comprising:
an audio input module;
an audio detection module in communication with the audio input module;
a speech detection module in communication with the audio detection module;
a wakeword recognition module in communication with the speech detection module; and
a network interface module in communication with the wakeword recognition module,
wherein:
the audio detection module is configured to:
receive audio input from the audio input module;
determine a volume of at least a portion of the audio input;
cause the audio input module to increase a sampling rate of the audio input based at least in part on the volume exceeding a threshold; and
cause activation of the speech detection module based at least in part on the volume exceeding the threshold;

the speech detection module is configured to determine a first score indicating a likelihood that the audio input comprises speech and cause activation of the wakeword recognition module based at least on part on the score; and
the wakeword recognition module is configured to:
determine a second score indicating a likelihood that the audio input comprises a wakeword; and
cause activation of a network interface module based on the second score by providing power to the network interface module; and

the network interface module is configured to transmit at least a portion of the obtained audio input to a remote computing device.

2. The system of claim 1 , wherein the audio input device comprises a microphone, the audio detection module comprises a first digital signal processor, the speech detection module comprises a second digital signal processor, and the wakeword recognition module comprises a microprocessor.

3. The system of claim 1 , wherein:
the speech detection module is further configured to determine the first score using at least one of a hidden Markov model, a Gaussian mixture model, energies in a plurality of spectral bands, or signal to noise ratios in a plurality of spectral bands; and
the wakeword recognition module is further configured to determine the second score using at least one of an application processing module, a hidden Markov model, and a Gaussian mixture model.

4. The system of claim 1 , wherein:
the wakeword recognition module is further configured to cause deactivation of the audio detection module based at least in part on the first score; and
the wakeword recognition module is further configured to cause deactivation of the speech detection module based at least in part on the second score.

5. A computer-implemented method of operating a first computing device, the method comprising:
receiving an audio input;
determining one or more values from the audio input, wherein the one or more values comprise at least one of:
a first value indicating an energy level of the audio input; or
a second value indicating a likelihood that the audio input comprises speech;

increasing a sampling rate of the audio input, from a first lower sampling rate to a second higher sampling rate, based at least in part on the one or more values;
activating a first module of the first computing device based at least in part on the one or more values;
performing an operation, by the first module, wherein the operation comprises at least one of:
determining that the audio input comprises a wakeword and causing activation of a network interface module in response to determining that the audio input comprises a wakeword, wherein causing activation of the network interface module comprises providing power to the network interface module;
performing speech recognition on at least a portion of the audio input to obtain speech recognition results; or
causing transmission of at least a portion of the audio input to a second computing device.

6. The computer-implemented method of claim 5 , wherein:
the first module comprises a processor that is switchable between a low-power state and a high-power state; and
the processor only performs the operation when it is in the high-power state.

7. The computer-implemented method of claim 6 , wherein activating the first module comprises switching the processor from the low-power state to the high-power state.

8. The computer-implemented method of claim 6 further comprising deactivating the first module, wherein deactivating the first module comprises switching the processor from the high-power state to the low-power state.

9. The computer-implemented method of claim 6 , wherein the processor comprises at least one of a digital signal processor or a microprocessor.

10. The computer-implemented method of claim 5 , wherein the first module comprises a software module configured to be executed by a microprocessor.

11. The computer-implemented method of claim 10 , wherein activating the first module comprises causing the microprocessor to execute the software module.

12. The computer-implemented method of claim 5 , wherein the operation further comprising receiving speech recognition results from the second computing device.

13. The computer-implemented method of claim 12 , wherein the speech recognition results comprise at least one of a transcription of at least a portion of the audio input and a response to an intelligent agent query included in at least a portion of the audio input.

14. The computer-implemented method of claim 12 further comprising:
activating a second module of the first computing device based at least in part on the one or more values, wherein the second module is configured to implement a speech recognition application; and
processing the speech recognition results with the speech recognition application.

15. The computer-implemented method of claim 5 , wherein providing power to the network interface module causes the network interface module to transition from a deactivated state to an activated state.

16. The computer-implemented method of claim 15 , wherein communications sent via the network interface module are prevented while the network interface module is in the deactivated state.

17. The computer-implemented method of claim 15 , wherein communications sent via the network interface module are enabled while the network interface module is in the activated state.

18. The computer-implemented method of claim 5 , wherein providing power to the network interface module comprises providing power to a processor of the first computing device.

19. The computer-implemented method of claim 5 , further comprising determining that the energy level of the audio input satisfies a threshold, wherein the increasing the sampling rate is performed in response to determining that the energy level of the audio input satisfies the threshold.

20. A device comprising:
a first processor configured to:
determine one or more values, wherein the one or more values comprise at least one of a first value indicating an energy level of an audio input or a second value indicating a likelihood that the audio input comprises speech; and
cause an increase in a sampling rate of the audio input, from a first lower sampling rate to a second higher sampling rate, based at least in part on the one or more values;
cause activation of a second processor based at least in part on the one or more values;

the second processor configured to perform an operation, wherein the operation comprises at least one of:
determining that the audio input comprises a wakeword and causing activation of a network interface module in response to determining that the audio inpu

Assignees

• Amazon Tech Inc

Inventors

• Basye Kenneth John
• Secker-Walker Hugh Evan
• David Tony
• Kneser Reinhard
• Adams Jeffrey Penrod
• Salvador Stan Weidner
• Krishnamoorthy Mahesh

Classifications

• G10L15/30

  Distributed recognition, e.g. in client-server systems, for mobile phones or network applications · CPC title

• G10L2015/088

  Word spotting · CPC title

• G10L15/28 Primary

  Constructional details of speech recognition systems · CPC title

• G10L25/78

  Detection of presence or absence of voice signals (switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10) · CPC title

• G10L15/32

  Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems · CPC title