Discriminative Training Based on the Criterion of Least Phone Competing Tokens for Large Vocabulary Speech Recognition.
Bo LiuHui JiangJian-Lai ZhouRen-Hua WangPublished in: ICASSP (1) (2005)
Keyphrases
- speech recognition
- acoustic models
- discriminative training
- automatic speech recognition
- hidden markov models
- speech recognizer
- speaker independent
- minimum classification error
- speech signal
- speech synthesis
- speaker identification
- language model
- pattern recognition
- noisy environments
- speech recognition systems
- broadcast news
- maximum likelihood
- speaker adaptation
- probabilistic interpretation
- information extraction
- support vector machine
- probabilistic model
- prior knowledge