Phoneme background model for information bottleneck based speaker diarization.
Sree Harsha YellaPetr MotlícekHervé BourlardPublished in: INTERSPEECH (2014)
Keyphrases
- information bottleneck
- speaker diarization
- background model
- speech recognition
- background subtraction
- information theoretic
- automatic speech recognition
- moving objects
- foreground objects
- broadcast news
- hidden markov models
- language model
- pattern recognition
- gaussian mixture model
- speech signal
- mutual information
- pixel wise
- segmentation algorithm
- speaker identification
- image sequences
- video sequences
- object detection
- image analysis
- graph cuts
- feature vectors