This thesis describes the design and implementation of a program voice input system. The system is composed of five stages : parameter extraction, segmentation, coarse classification of the words, word recognition, and syntactic analysis. Fuzzy pattern matching technique is used to recognize the segmented words, any dynamic time warping technique is used to build the reference patterns. Finally, the recognition result is corrected in the syntactic analysis stage. A new method of syntactic analysis which uses the relative position of words in sentences is proposed.
The speech data were uttered by two men. The recognition rate in word recognition stage is about 93%, but the correct rate grows up to 98% after the syntactic analysis stage.