Hybrid model method for automatic segmentation of mandarin TTS corpus

For a corpus-based Mandarin text-to-speech system, the quality of synthesized speech is highly affected by the accuracy of unit boundaries. In this paper, we proposed a hybrid model method for automatic segmentation of Mandarin text-to-speech corpus. The boundaries of acoustic units are categorized into eleven phonetic groups. For a given phonetic group of boundaries, the proposed method selects an appropriate model from initial-final monophone-based HMM, semi-syllable monophone-based HMM and initial-final triphone-based HMM. The experimental results show that the hybrid model method can achieve better performance than the single model method, in terms of error rate and time shift of boundaries.