Low-power, always-listening, voice command detection and capture

Patent metadata

Field	Value
Publication number	US-10403279-B2
Application number	US-201715706178-A
Country	US
Kind code	B2
Filing date	Sep 15, 2017
Priority date	Dec 21, 2016
Publication date	Sep 3, 2019
Grant date	Sep 3, 2019

Abstract

A system for detecting and capturing voice commands, the system comprising a voice-activity detector (VAD) configured to receive a VAD-received digital-audio signal; determine the amplitude of the VAD-received digital-audio signal; compare the amplitude of the VAD-received digital-audio signal to a first threshold and to a second threshold; withhold a VAD interrupt signal when the amplitude of the VAD-received digital-audio signal does not exceed the first threshold or the second threshold; generate the VAD interrupt signal when the amplitude of the VAD-received digital-audio signal exceeds the first threshold and the second threshold; and perform spectral analysis of the VAD-received digital-audio signal when the amplitude of the VAD-received digital-audio signal is between the first threshold and the second threshold.

First claim

The invention claimed is:

1. A system for detecting and capturing voice commands, the system comprising:
an input configured to receive an audio signal;
an analog to digital converter configured to convert the audio signal to a digital-audio signal;
an amplitude detector configured determine the amplitude of the digital-audio signal;
a controller configured to compare the amplitude of the digital-audio signal to a first threshold and to a second threshold and withhold an interrupt signal when the amplitude of the digital-audio signal does not exceed the first threshold or the second threshold, generate the interrupt signal when the amplitude of the digital-audio signal exceeds the first threshold and the second threshold, and trigger spectral analysis of the digital-audio signal when the amplitude of the digital-audio signal is between the first threshold and the second threshold.

2. The system of claim 1 in which the analog to digital converter includes a wakeup circuit configured to:
determine whether the audio signal has an onset level of audio activity;
when the audio signal does not have the onset level of audio activity, place the analog to digital converter in a sleep mode in which a converter portion of the analog to digital converter does not generate the digital-audio signal; and
when the audio signal has the onset level of audio activity, place the analog to digital converter in an active mode in which the converter portion of the analog to digital converter generates the digital-audio signal.

3. The system of claim 2 in which, to determine whether the an audio signal has an onset level of audio activity, the wakeup circuit is configured to compare the audio signal to a wakeup-onset threshold.

4. The system of claim 2 in which the wakeup circuit comprises:
an audio preamp a gain stage configured to boost the audio signal; and
a comparator stage configured to compare the boosted audio signal to a wakeup-onset threshold indicating the onset level of audio activity.

5. The system of claim 2 in which the converter portion of the analog to digital converter is unpowered when the analog to digital converter is in the sleep mode, and in which the converter portion of the analog to digital converter is powered when the analog to digital converter in the active mode.

6. The system of claim 2 further comprising a decimation filter between the analog to digital converter and the amplitude detector, the decimation filter configured to receive the digital-audio signal and process the digital-audio signal to generate reduced digital-audio samples having a reduced sampling rate, the reduced sampling rate being lower than the analog to digital converter sampling rate.

7. The system of claim 1 further comprising:
a sample combine block configured to receive as an input the digital-audio signal having a sampling rate and to process the digital-audio signal to generate a combined digital-audio signal having an amplitude, the combined digital-audio signal having a sampling rate lower than the sampling rate of the digital-audio signal.

8. The system of claim 1 further comprising a digital signal processor configured to receive the interrupt signal and analyze the digital-audio signal for a trigger phrase.

9. The system of claim 8 further comprising a voice random-access memory configured to receive and store the digital-audio signal in a first-in, first-out format, the voice random-access memory being configured to be accessed by the digital signal processor.

10. The system of claim 8 further comprising a clock system comprising a plurality of asynchronous clock generating mechanisms, each of the asynchronous clock generating mechanisms having a different current consumption, the clock system being configured to switch between the asynchronous clock generating mechanisms.

11. A method of detecting and capturing voice commands comprising:
receiving, by a voice-activity detector, a digital-audio signal;
determining, by the voice-activity detector, the amplitude of the digital-audio signal;
comparing, by the voice-activity detector, the amplitude of the digital-audio signal to a first threshold and to a second threshold;
withholding, by the voice-activity detector, an interrupt signal when the amplitude of the digital-audio signal does not exceed the first threshold or the second threshold;
generating, by the voice-activity detector, the interrupt signal when the amplitude of the digital-audio signal exceeds the first threshold and the second threshold; and
performing, by the voice-activity detector, spectral analysis of the digital-audio signal when the amplitude of the digital-audio signal is between the first threshold and the second threshold.

12. The method of claim 11 further comprising:
receiving, by an analog-to-digital converter, an audio signal;
determining, by a wakeup circuit, whether the audio signal has an onset level of audio activity; and
when the audio signal does not have the onset level of audio activity, placing the analog to digital converter in a sleep mode in which the converter portion of the analog to digital converter does not generate a digital-audio signal; and
when the audio signal has the onset level of audio activity, placing the analog to digital converter in an active mode in which a converter portion of the analog to digital converter generates a digital-audio signal from the audio signal at an analog to digital converter sampling rate.

13. The method of claim 12 in which determining, by a wakeup circuit, whether the audio signal has an onset level of audio activity comprises comparing the audio signal to a wakeup-onset threshold.

14. The method of claim 12 further comprising:
unpowering the converter portion of the analog to digital converter when the analog to digital converter in the sleep mode; and
powering the converter portion of the analog to digital converter when the analog to digital converter in the active mode.

15. The method of claim 12 further comprising:
receiving, by a digital signal processor, the interrupt signal from the voice-activity detector; and
analyzing, by the digital signal processor, the digital-audio signal to detect a trigger phrase.

16. The method of claim 15 further comprising receiving and storing, by a voice random-access memory, the digital-audio signal in a first-in, first-out format; and, when a trigger phrase is detected by the digital signal processor, discarding all data on the voice RAM before an endpoint of the trigger phrase.

17. The method of claim 15 further comprising, when a trigger phrase is detected by the digital signal processor, determining whether a user spoke during a prespecified timeout period after the trigger phrase.

18. A voice-activity detector, comprising:
an input configured to receive a digital-audio signal;
an amplitude detector configured determine the amplitude of the digital-audio signal; and
a controller configured to withhold an interrupt signal when the amplitude of the digital-audio signal does not exceed a first threshold or a second threshold, generate the interrupt signal when the amplitude of the digital-audio signal exceeds the first threshold and the second threshold, and trigger spectral analysis of the digital-audio signal when the amplitude of the digital-audio signal is between the first threshold and the second threshold; and
a spectral analyzer configured to perform the spectral analysis of the digital-audio signal to instruct the controller whether to generate the interrupt signal.

19.

Assignees

• Avnera Corp

Inventors

• Zhao Xudong
• Stange Alexander C
• O'Connor Shawn
• Hadiashar Ali

Classifications

• G10L25/78 Primary

  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

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G10L2025/783

  based on threshold decision · CPC title

• G10L2015/088

  Word spotting · CPC title

• G10L25/84

  for discriminating voice from noise · CPC title