Modeling spectral envelopes using deep conditional restricted Boltzmann machines for statistical parametric speech synthesis.
Xiang YinZhen-Hua LingYa-Jun HuLi-Rong DaiPublished in: ICASSP (2016)
Keyphrases
- restricted boltzmann machine
- speech synthesis
- deep learning
- deep belief networks
- speech recognition
- text to speech
- conditional random fields
- probabilistic graphical models
- vocal tract
- contrastive divergence
- machine learning
- information retrieval
- learning algorithm
- bayesian networks
- pairwise
- dynamic bayesian networks