Abstract: The paper reports on a simulation system which can synthesize unlimited Chinese vocabulary using diphones as sound elements according to the principle of formant vocoder. A spelt Chinese text can directly be used as its input without special transcription for pronunciation and its sound output is standard Chinese speech （manderin）. A general method for synthetic speech, such as the processing method of excite source （voiced, unvoiced） and the terminal analog of vocal system, has been introduced in the paper. A proximat method of 4 tones for Chinese words and a segment library made of about 640 diphones for synthesis of unlimited Chinese vocabulary have been also discripted according to some features of Chinese language, for example, there are a lot of homophones, the assimilation of some phonemes, the function of recognizing different meaning from different pitch contour in chinese language, and so on.
A text about 400 Chinese words has been synthsized on the simulation system and has been dictated to 10 Chinese people. Average sentence intelligibility is near 90%. The best, 97%.