Language-invariant Bottleneck Features from Adversarial End-to-end Acoustic Models for Low Resource Speech Recognition.
Jiangyan YiJianhua TaoYe BaiPublished in: ICASSP (2019)
Keyphrases
- speech recognition
- end to end
- acoustic models
- speech recognition systems
- hidden markov models
- speech recognizer
- automatic speech recognition
- speaker independent
- language model
- speech signal
- feature space
- feature extraction
- feature vectors
- feature set
- pattern recognition
- noisy environments
- low level
- natural language
- speaker identification
- speaker diarization
- maximum likelihood
- discriminative training
- information retrieval
- probabilistic model
- image processing