Back to results
US 8,078,463 B2Granted
Method and apparatus for speaker spotting
AssigneeNice Systems Ltd.
InventorsMoshe Wasserblat, Yaniv Zigel, Oren Pereg
Life of this patent
Estimating the term…
Abstract
A method and apparatus for spotting a target speaker within a call interaction by generating speaker models based on one or more speaker's speech; and by searching for speaker models associated with one or more target speaker speech files.
Claims38
A computerized method for spotting an at least one call interaction out of a multiplicity of call interactions, in which an at least one target speaker participates, the method comprising:
- capturing at least one target speaker speech sample of the at least one target speaker by a speech capture device;
- generating by a computerized engine a multiplicity of speaker models based on a multiplicity of speaker speech samples from the at least one call interaction;
- matching by a computerized server the at least one target speaker speech sample with speaker models the multiplicity of speaker models to determine a target speaker model;
- determining a score for each call interaction of the multiplicity of call interactions according to a comparison between the target speaker model and the multiplicity of speaker models;
- and
- based on scores that are higher than a predetermined threshold, determining call interactions, of the multiplicity of call interactions, in which the at least one target speaker participates.
3 more elements
A computerized apparatus for spotting an at least one call interaction out of a multiplicity of call interactions in which a target speaker participates, the apparatus comprising:
- a training computerized component configured for generating a multiplicity of speaker models based on a multiplicity of speaker speech samples from the at least one call interaction;
- and
- a speaker spotting computerized component configured for matching the target speaker speech sample with speaker models of the multiplicity of speaker models to determine a target speaker model,
- determining a score for each call interaction of the multiplicity of call interactions according to a comparison between the target speaker model and the multiplicity of speaker models, and
- based on scores that are higher than a predetermined threshold, determining call interactions, of the multiplicity of call interactions, and in which the at least one target speaker participates.
2 more elements
Dependencies are inferred from the claim text.