Segment-based speaker verification using dynamically generated phrases

A computer-implemented method comprising:

• identifying candidate enrollment phrases to enroll a particular user for voice verification, each candidate enrollment phrase includes at least one subword and at least one candidate enrollment phrase includes at least one subword for which no stored enrollment audio data representing the user speaking the subword has been obtained;
• prompting the particular user to speak candidate phrases including the at least one candidate enrollment phrase that contains at least one subword for which no stored enrollment audio data representing the particular user speaking the subword has been obtained;
• obtaining and storing enrollment audio data representing the particular user speaking the candidate enrollment phrases until enrollment audio data has been obtained that meets a certain threshold;
• dynamically generating a verification phrase based at least on one or more of the subwords included in the candidate enrollment phrases uttered by the particular user in the enrollment acoustic data;
• prompting a user to speak the dynamically generated verification phrase;
• obtaining verification audio data representing the user speaking the dynamically generated verification phrase;
• comparing the obtained verification audio data with the enrollment verification audio data to determine whether the user speaking the dynamically generated verification phrase is the particular user who spoke the candidate enrollment phrases;
• and
• in response to determining that the user speaking the dynamically generated verification phrase is the particular user who spoke the candidate enrollment phrases, verifying an identity of the user as the particular user.

A system comprising:

• one or more data processing apparatuses;
• and
• one or more storage devices storing instructions that are operable, when executed by one or more data processing apparatuses, to cause the one or more data processing apparatuses to perform operations comprising:
• identifying candidate enrollment phrases to enroll a particular user for voice verification, each candidate enrollment phrase includes at least one subword and at least one candidate enrollment phrase includes at least one subword for which no stored enrollment audio data representing the user speaking the subword has been obtained;
• prompting the particular user to speak candidate phrases including the at least one candidate enrollment phrase that contains at least one subword for which no stored enrollment audio data representing the particular user speaking the subword has been obtained;
• obtaining and storing enrollment audio data representing the particular user speaking the candidate enrollment phrases until enrollment audio data has been obtained that meets a certain threshold;
• dynamically generating a verification phrase based at least on one or more of the subwords included in the candidate enrollment phrases uttered by the particular user in the enrollment acoustic data;
• prompting a user to speak the dynamically generated verification phrase;
• obtaining verification audio data representing the user speaking the dynamically generated verification phrase;
• comparing the obtained verification audio data with the enrollment verification audio data to determine whether the user speaking the dynamically generated verification phrase is the particular user who spoke the candidate enrollment phrases;
• and
• in response to determining that the user speaking the dynamically generated verification phrase is the particular user who spoke the candidate enrollment phrases, verifying an identity of the user as the particular user.

One or more non-transitory computer-readable storage mediums storing instructions thereon that are executable by a processing device and upon such execution cause the processing device to perform operations comprising:

• identifying candidate enrollment phrases to enroll a particular user for voice verification, each candidate enrollment phrase includes at least one subword and at least one candidate enrollment phrase includes at least one subword for which no stored enrollment audio data representing the user speaking the subword has been obtained;
• prompting the particular user to speak candidate phrases including the at least one candidate enrollment phrase that contains at least one subword for which no stored enrollment audio data representing the particular user speaking the subword has been obtained;
• obtaining and storing enrollment audio data representing the particular user speaking the candidate enrollment phrases until enrollment audio data has been obtained that meets a certain threshold;
• dynamically generating a verification phrase based at least on one or more of the subwords included in the candidate enrollment phrases uttered by the particular user in the enrollment acoustic data;
• prompting a user to speak the dynamically generated verification phrase;
• obtaining verification audio data representing the user speaking the dynamically generated verification phrase;
• comparing the obtained verification audio data with the enrollment verification audio data to determine whether the user speaking the dynamically generated verification phrase is the particular user who spoke the candidate enrollment phrases;
• and
• in response to determining that the user speaking the dynamically generated verification phrase is the particular user who spoke the candidate enrollment phrases, verifying an identity of the user as the particular user.