Application of eigenvoice technique to spectrum and pitch pattern modeling in HMM-based speech synthesis

摘要

This paper describes an HMM-based speech synthesis system which can generate various voice qualities by using the eigenvoice technique. The eigenvoice technique, in which a large number of speaker-dependent HMM sets are represented by a few parameters through a dimensionality reduction. technique, e.g., PCA, has been used for speaker adaptation in HMM-based speech recognition. We apply the eigenvoice technique to an HMM-based speech synthesis system in which spectrum, and pitch are modeled by HMM's simultaneously. The generated spectrum and pitch patterns shown, and some discussion on the voice qualities of the synthesised speech is given