Prosodic model makes the speech synthesis system has a rhythm of the simulation and emulation capabilities, and it plays a crucial role to enhance the naturalness of synthesized speech. The paper studies Tibetan rhythmic structure based on the characteristics of the Lhasa Tibetan, and determines the parameters that affect the prosodic features. To meet the actual needs of establishing high-quality Tibetan prosodic model, the paper designs corpus selection rules and optimizes it. Lhasa Tibetan annotation system is proposed and the Lhasa Tibetan prosodic model corpus has been eventually established. The corpus has highly and scientifically completed, which basically covers Lhasa Tibetan segment and suprasegmental information. It will play a huge role in promoting the study of Tibetan prosodic model and in improving the naturalness of synthesized speech.
展开▼