A Novel Iterative Speaker Model Alignment Method from Non-Parallel Speech for Voice Conversion.
Peng SongWenming ZhengXinran ZhangYun JinCheng ZhaMinghai XinPublished in: IEICE Trans. Fundam. Electron. Commun. Comput. Sci. (2015)
Keyphrases
- mathematical model
- statistical model
- modeling method
- probabilistic model
- evaluation method
- theoretical analysis
- objective function
- prior knowledge
- cost function
- prediction model
- prior information
- hybrid method
- linear model
- input data
- speaker recognition
- parallel implementation
- speech recognition
- parameter estimation
- markov random field
- bp neural network
- gaussian distribution
- segmentation method
- em algorithm
- optimization method
- iterative process
- speech synthesis
- neural network
- automatic speech recognition
- test data
- energy function
- support vector machine
- similarity measure