Machine learning for improving quality of voice biometrics
AssigneeCapital One Services, LLC
InventorsBozhao Tan, Isabelle Alice Yvonne Moulinier, David Almquist, June Wu
Life of this patent
Estimating the term…
Abstract
Methods and systems are disclosed herein for improving the quality of audio for use in a biometric. A biometric system may use machine learning to determine whether audio or a portion of the audio should be used as a biometric for a user. A sample of the user's voice may be used to generate a voice signature of the user. Portions of the audio that do not meet a similarity threshold when compared with the voice signature may be removed from the audio. Additionally or alternatively, interfering noises may be detected and removed from the audio to improve the quality of a voice biometric generated from the audio.
Claims20
A system for identifying which segments of a phone call recording may be used as a biometric sample to authenticate a user, the system comprising:
- one or more processors and computer program instructions that, when executed, cause the one or more processors to perform operations comprising:
- receiving an audio file of a phone call between a user and one or more call agents, wherein the audio file comprises a first portion of the phone call and a second portion of the phone call, and wherein the first portion comprises voice data associated with a first call agent and the second portion comprises voice data associated with a user and a second call agent to whom the phone call has been transferred from the first call agent;
- dividing the second portion of the phone call into a plurality of segments comprising a first subset of segments and a second subset of segments, wherein the first subset of segments includes one or more beginning segments of the second portion of the phone call;
- generating, based on inputting the first subset of segments into a machine learning model, a signature vector indicative of voice audio of the user;
- generating, based on inputting the second subset of segments into the machine learning model, a plurality of vectors comprising a vector for each segment of the second subset of segments;
- comparing the signature vector with each of the plurality of vectors for the second subset of segments;
- determining, based on the comparing, that the signature vector does not match one or more vectors of the plurality of vectors;
- in response to determining that the signature vector does not match the one or more vectors of the plurality of vectors, removing, from the second portion of the phone call, one or more segments corresponding to the one or more vectors that do not match the signature vector;
- generating a voice biometric for the user using the second portion of the phone call without the one or more segments;
- and
- providing the voice biometric to a voice based user authentication system.
A method comprising:
- receiving, by a computing device, of a call comprising a first portion corresponding to an agent and a second portion corresponding to a user;
- dividing the second portion into a plurality of segments comprising a first subset of segments and a second subset of segments;
- generating, based on the first subset of segments, a signature vector indicative of voice audio of the user;
- generating a plurality of vectors comprising a vector for each segment of the second subset of segments;
- comparing the signature vector with each of the plurality of vectors for the second subset of segments;
- determining, based on the comparing, that the signature vector does not match one or more vectors of the plurality of vectors;
- in response to the determining that the signature vector does not match the one or more vectors of the plurality of vectors, removing, from the second portion, one or more segments corresponding to the one or more vectors that do not match the signature vector;
- generating a voice biometric for the user using the second portion without the one or more segments;
- and
- providing the audio signature to a voice based user authentication system.
A tangible, non-transitory, machine-readable medium storing instructions that when executed by one or more processors effectuate operations comprising:
- receiving of a call comprising a first portion corresponding to an agent and a second portion corresponding to a user;
- dividing the second portion into a plurality of segments comprising a first subset of segments and a second subset of segments;
- generating, based on the first subset of segments, a signature vector indicative of voice audio of the user;
- generating a plurality of vectors comprising a vector for each segment the second subset of segments;
- comparing the signature vector with each of the plurality of vectors for the second subset of segments;
- determining, based on the comparing, that the signature vector does not match one or more vectors of the plurality of vectors;
- in response to the determining that the signature vector does not match the one or more vectors of the plurality of vectors, removing, from the second portion, one or more segments corresponding to the one or more vectors that do not match the signature vector;
- generating a voice biometric for the user using the second portion without the one or more segments;
- and
- providing the audio signature to a voice based user authentication system.
Dependencies are inferred from the claim text.