Automatic optimization of data perturbation distributions for multi-style training in speech recognition.
Mortaza DoulatyRichard RoseOlivier SiohanPublished in: SLT (2016)
Keyphrases
- speech recognition
- wall street journal corpus
- data perturbation
- isolated word
- hidden markov models
- language model
- speech synthesis
- acoustic models
- automatic speech recognition
- speech processing
- speech recognizer
- pattern recognition
- noisy environments
- privacy preserving data mining
- speech recognition technology
- speech signal
- privacy preserving
- speaker identification
- statistical databases
- speech recognition systems
- speaker independent
- text classification
- speaker dependent