A Solution on Tibetan Speech Synthesis Based on HMM

Speech synthesis is mainly utilized to solve the problem of transforming text information to speech information. For the Tibetan speech synthesis, this paper proposes a complete solution on speech synthesis adapting the rules of the Tibetan language. In the proposed method, the key technologies on the front and back ends of the synthesis are respectively elaborated, including speech corpus selecting, Tibetan-Latin transliterating, word segmentation, text analyzing, prosodic labeling, vocoder technology, speech modeling, and problem-set design, etc. The experiment results illustrate that the test system of Tibetan speech synthesis using the proposed methods can obtain good performances.