Unsupervised Speaker Adaptation Using All-Phoneme Ergodic Hidden Markov Network.
Yasunaga MiyazawaJun-ichi TakamiShigeki SagayamaShoichi MatsunagaPublished in: IEICE Trans. Inf. Syst. (1995)
Keyphrases
- markov networks
- speaker adaptation
- maximum likelihood
- speaker dependent
- speech recognition
- graphical models
- em algorithm
- automatic speech recognition
- semi supervised
- maximum a posteriori
- unsupervised learning
- mixture model
- probabilistic model
- first order logic
- supervised learning
- belief propagation
- bayesian inference
- speaker independent
- bayesian networks
- machine learning
- markov random field
- face recognition
- speech recognition systems
- feature vectors
- maximum entropy
- conditional probabilities
- hidden markov models
- gaussian mixture model
- conditional random fields
- low level
- neural network