Unsupervised Style and Content Separation by Minimizing Mutual Information for Speech Synthesis.
Ting-Yao HuAshish ShrivastavaOncel TuzelChandra DhirPublished in: ICASSP (2020)
Keyphrases
- speech synthesis
- mutual information
- speech recognition
- text to speech
- image registration
- similarity measure
- information theoretic
- prosodic features
- medical image registration
- multi modal image registration
- vocal tract
- unsupervised learning
- feature selection
- mutual information maximization
- data driven
- information gain
- information theoretic measures
- metadata
- competitive learning
- supervised learning
- semi supervised
- web content
- multimedia content
- user generated content
- image coding
- pattern recognition
- speech corpus
- information retrieval