Bottleneck linear transformation network adaptation for speaker adaptive training-based hybrid DNN-HMM speech recognizer.
Tsubasa OchiaiShigeki MatsudaHideyuki WatanabeXugang LuHisashi KawaiShigeru KatagiriPublished in: ICASSP (2016)
Keyphrases
- speaker adaptation
- speech recognition
- speech recognizer
- linear transformation
- acoustic models
- hidden markov models
- training process
- speaker independent
- automatic speech recognition
- spoken language
- broadcast news
- speaker dependent
- language model
- pattern recognition
- speech signal
- speaker identification
- metric learning
- unsupervised learning
- model selection
- dimensionality reduction
- active learning
- query processing
- information retrieval
- neural network