The prominence analysis and synthesis of emphasis in Putonghua

Emphasis is an important feature of intonation. The technology of emphatic speech synthesis can improve the naturalness and the expressiveness of the synthesized speech. This paper defined the prominences of acoustic features, analyzed the prominences of the acoustic features of the emphasized syllables in different prosody locations, e.g. the head, the body, and found that the prominences the maximum F0 of the emphasized syllables at the end of prosody units(prosody words or prosody phrases) are lower than other emphasized syllables. A parameter generation algorithm of emphasized syllables based on the prominences of the acoustic features was proposed based on the analysis. It avoided the problem of oversize modification of the traditional linear modification algorithm. An emphatic speech synthesis system based on hidden Markov model (HMM) was built with the proposed algorithm. Experiments demonstrated that the system could synthesize emphasized speeches and improve the naturalness and the expressive of the synthesized speeches.