Rapid unsupervised speaker adaptation based on multi-template HMM sufficient statistics in noisy environments.
Randy GomezAkinobu LeeHiroshi SaruwatariKiyohiro ShikanoPublished in: INTERSPEECH (2005)
Keyphrases
- speaker adaptation
- noisy environments
- speech recognition
- automatic speech recognition
- sufficient statistics
- hidden markov models
- speech signal
- speaker independent
- speech recognizer
- digit recognition
- speaker identification
- speaker verification
- language model
- maximum likelihood
- speech enhancement
- unsupervised learning
- speech synthesis
- pattern recognition
- supervised learning
- broadcast news
- markov models
- noise reduction
- speech recognition systems
- maximum entropy
- semi supervised
- k means
- image segmentation