Integrating global variance of log power spectrum derived from LSPs into MGE training for HMM-based parametric speech synthesis.
Yu-Sheng SunZhen-Hua LingXiang YinLi-Rong DaiPublished in: ISCSLP (2014)
Keyphrases
- power spectrum
- speech synthesis
- speech recognition
- frequency domain
- text to speech
- natural images
- hidden markov models
- frequency analysis
- vocal tract
- prosodic features
- power spectral
- signal noise ratio
- training set
- autocorrelation function
- density function
- computer vision
- correlation coefficient
- neural network
- image quality
- higher order
- information retrieval