With the expected and continued trend toward friendlier man-machine interfaces, the applications and demands for technology of synthesis of natural-sounding are growing rapidly. In this paper, a Chinese Text-to-Speech system is developed, in which some speech input modes of Chinese character, such as keyboard input, on-line handwritten Chinese character input, etc., are designed. Synthesizing speech form of 3755 Chinese characters, which are commonly used, included in the first level China National Standard of basic set of Chinese characters, arc pro. duced. By applying the method of multi-pulse linear predictive analysis-by-syn. thesis, speech feature parameters arc determined, and a phonetic database, which includes all Chinese syllables, has been established. As all the syllables have been elaborately regulated, high-quality speech is produced. Meanwhile, a Chinese phonetic table is built up to check the index of the database. Besides, an improved algorithm of multi-pulse linear predictive coding is proposed. Compared with tranditional algorithms, speech quality can be significantly improved by using this method.
修改评论