Combining Adversarial Training and Disentangled Speech Representation for Robust Zero-Resource Subword Modeling.
Siyuan FengTan LeeZhiyuan PengPublished in: INTERSPEECH (2019)
Keyphrases
- speech recognition
- model construction
- broadcast news
- resource allocation
- training set
- noisy environments
- n gram
- training examples
- computationally efficient
- test set
- image representation
- modeling tool
- acoustic models
- training algorithm
- audio visual
- automatic speech recognition
- online learning
- pattern recognition
- spoken document retrieval