Training speaker recognition models for digital assistants

  • US 10,789,959 B2
  • Filed: 06/04/2018
  • Issued: 09/29/2020
  • Est. Priority Date: 03/02/2018
  • Status: Active Grant
  • ×
    • Pin Icon | RPX Insight
    • Pin
First Claim
Patent Images

1. A non-transitory computer-readable storage medium storing one or more programs configured to be executed by one or more processors of an electronic device, the one or more programs including instructions for:

  • obtaining, at a first time, user authentication information;

    receiving, at a second time, a user utterance representing a request to perform one or more tasks;

    generating a voice print from the user utterance;

    determining whether a plurality of conditions are satisfied, the plurality of conditions including;

    a first condition that the user authentication information corresponds to one or more authentication credentials stored in the memory, wherein the one or more authentication credentials are assigned to a registered user of the electronic device; and

    a second condition that the first time and the second time are not separated by more than a predefined time period;

    in accordance with determining that the plurality of conditions are satisfied, updating, based on the voice print, a speaker profile assigned to the registered user,after updating the speaker profile, determining whether a total number of voice prints associated with the speaker profile is equal to a predefined number; and

    in accordance with determining that the total number of voice prints associated with the speaker profile is equal to a predefined number, providing an output prompting a user to activate a voice invocation feature of the digital assistant.

View all claims