Unsupervised speaker adaptation of BLSTM-RNN for LVCSR based on speaker code.
Zhiying HuangShaofei XueZhijie YanLi-Rong DaiPublished in: ISCSLP (2016)
Keyphrases
- speaker adaptation
- speech recognition
- maximum likelihood
- speaker dependent
- automatic speech recognition
- recurrent neural networks
- nearest neighbor
- speech recognizer
- speaker independent
- hidden markov models
- speaker identification
- vocal tract
- unsupervised learning
- broadcast news
- speech synthesis
- language model
- pattern recognition
- image processing
- spoken term detection
- speech signal
- visual information
- multi modal