Back to results
US 8,554,562 B2Granted
Method and system for speaker diarization
AssigneeNuance Communications, Inc.
InventorHagai Aronowitz
Life of this patent
Estimating the term…
Abstract
A method and system for speaker diarization are provided. Pre-trained acoustic models of individual speaker and/or groups of speakers are obtained. Speech data with multiple speakers is received and divided into frames. For a frame, an acoustic feature vector is determined extended to include log-likelihood ratios of the pre-trained models in relation to a background population model. The extended acoustic feature vector is used in segmentation and clustering algorithms.
Claims15
A computer program product for speaker diarization, the computer program product comprising:
- a non-transitory computer readable medium;
- computer program instructions operative to:
- obtain pre-trained acoustic models of individual speakers and/or groups of speakers;
- receive speech data with multiple speakers;
- divide the speech data into frames;
- for each of a plurality of frames, calculate for one or more of the pre-trained acoustic models, a probability that the speaker of the frame is the speaker of the pre-trained acoustic model;
- for each of the plurality of frames, determine an acoustic feature vector representing the frame;
- for each of the plurality of frames, modify the respective acoustic feature vector of the frame to include one or more elements representing the calculated respective one or more probabilities;
- and
- segmenting or clustering the received speech using the modified feature vectors of the plurality of frames,
- wherein said program instructions are stored on said computer readable medium.
8 more elements
A system for speaker diarization, comprising:
- a processor;
- a storage medium storing pre-trained acoustic models of individual speakers and/or groups of speakers;
- a receiver for receiving speech data from multiple speakers;
- and
- a computer processor configured to divide the speech data into frames,
- determine for each of a plurality of frames a modified acoustic feature vector including elements representing for one or more of the pre-trained acoustic models, a probability that the speaker of the frame is the speaker of the pre-trained acoustic model, and segment or cluster the speech data using the modified feature vectors of the plurality of frames.
3 more elements
Dependencies are inferred from the claim text.