×

Speech endpointing based on word comparisons

  • US 10,140,975 B2
  • Filed: 05/17/2016
  • Issued: 11/27/2018
  • Est. Priority Date: 04/23/2014
  • Status: Active Grant
First Claim
Patent Images

1. A computer-implemented method comprising:

  • receiving, from a given user and by a microphone of a mobile device that includes (i) the microphone, (ii) an automated speech recognition system, and (iii) an end of utterance detector that is configured to identify an endpoint of an utterance spoken by a user in response to determining that a speaker has stopped speaking for a fixed duration, a first utterance;

    determining, by the end of utterance detector, that the given user has stopped speaking for the fixed duration after the first utterance;

    generating, by the automated speech recognition system, a first transcription of the first utterance;

    based on the first transcription of the first utterance, maintaining the microphone in an active state without endpointing the first utterance;

    after the given user has stopped speaking for at least the fixed duration after the first utterance, receiving, by the microphone and from the given user, a second utterance;

    generating, by the automated speech recognition system, a second transcription of the second utterance;

    based on both the first transcription and the second transcription, deactivating the microphone and endpointing the second utterance;

    in response to endpointing the second utterance, submitting, by the mobile device, a single search query that includes both the first transcription and the second transcription;

    receiving, by the mobile device, search results in response to the single search query that includes both the first transcription and the second transcription; and

    providing, for output by the mobile device, the search results.

View all claims
  • 2 Assignments
Timeline View
Assignment View
    ×
    ×