Back to results
US 11,776,547 B2Granted
System and method of video capture and search optimization for creating an acoustic voiceprint
AssigneeVerint Systems Ltd.
InventorsOmer Ziv, Ran Achituv, Ido Shapira, Jeremie Dreyfuss
Life of this patent
Estimating the term…
Abstract
Systems and method of diarization of audio files use an acoustic voiceprint model. A plurality of audio files are analyzed to arrive at an acoustic voiceprint model associated to an identified speaker. Metadata associate with an audio file is used to select an acoustic voiceprint model. The selected acoustic voiceprint model is applied in a diarization to identify audio data of the identified speaker.
Claims20
A method for creating a background acoustic voiceprint, the method comprising:
- receiving, by a processor, a plurality of audio files from an audio database server;
- diarizing, by the processor, each of the plurality of audio files by:
- separating the audio file into a plurality of speaker segments, wherein each speaker segment is a segment of speech separated by non-speech;
- clustering each of the speaker segments for the audio file into groups of similar speaker segments, wherein each group of similar speaker segments has a high likelihood of containing speech from a single speaker;
- and
- classifying each of the clustered groups of speaker segments as belonging to agents or as belonging to an other speaker, wherein the other speaker is not an agent;
- generating, by the processor, the background acoustic voiceprint based on the clustered speaker segments classified as belonging to the other speaker;
- receiving, by the processor, a new audio file;
- separating the new audio file into a new plurality of speaker segments, wherein each new speaker segment is a segment of speech separated by non-speech;
- applying, by the processor, the background acoustic voiceprint to each new speaker segment for the new audio file to determine which new speaker segments match the acoustic voiceprint;
- and
- labeling, by the processor, each new speaker segments of the new audio file that de does not match the acoustic voiceprint as belonging to an agent.
9 more elements
A system for creating a background acoustic voiceprint, the system comprising:
- an audio database server including a plurality of audio files;
- a processor;
- and
- a memory storing instructions, which when executed by the processor causes the processor to:
- receive a plurality of audio files from the audio database server;
- diarize each of the plurality of audio files by:
- separate the audio file into a plurality of speaker segments, wherein each speaker segment is a segment of speech separated by non-speech;
- cluster each of the speaker segments for the audio filed into groups of similar speaker segments, wherein each group of similar speaker segments has high likelihood of containing speech from a single speaker;
- and
- classify each of the clustered groups of speaker segments as belonging to agents or as belonging to an other speaker, wherein the other speaker is not an agent;
- generate the background acoustic voiceprint based on the clustered speaker segments classified as belonging to the other speaker;
- receive a new audio file;
- separate the new audio file into a new plurality of speaker segments, wherein each new speaker segment is a segment of speech separated by non-speech;
- apply the background acoustic voiceprint to each new speaker segment for the new audio file to determine which new speaker segments match the acoustic voiceprint;
- and
- label each new speaker segments of the new audio file that do not match the acoustic voiceprint as belonging to an agent.
13 more elements
A non-transitory computer-readable medium having instructions stored thereon for facilitating diarization of audio files, wherein the instructions, when executed by a processing system, direct the processing system to:
- receive a plurality of audio files from the audio database server;
- diarize each of the plurality of audio files by:
- separate the audio file into a plurality of speaker segments, wherein each speaker segment is a segment of speech separated by non-speech;
- cluster each of the speaker segments for the audio file into groups of similar speaker segments, wherein each group of similar speaker segments has a high likelihood of containing speech from a single speaker;
- and
- classify each of the clustered groups of speaker segments as belonging to agents or as belonging to an other speaker, wherein the other speaker is not an agent;
- build a background acoustic voiceprint based on the clustered speaker segments classified as belonging to the other speaker;
- receive a new audio file;
- separate the new audio file into a new plurality of speaker segments, wherein each new speaker segment is a segment of speech separated by non-speech;
- apply the background acoustic voiceprint to each new speaker segment for the new audio file to determine which new speaker segments match the acoustic voiceprint;
- and
- label each new speaker segments of the new audio file that does not match the acoustic voiceprint as belonging to an agent.
9 more elements
Dependencies are inferred from the claim text.