Audio-Stream Segment Base On Energy And Control By Multi-model Automat

Speech recognition, which combine digital signal process with phonetics to solve the information exchange problems between the men or man-machine in model communication domain, is a important branch of information science research in today, and make it be the last goal that computer could understand human speech.In 1990s, many practically research directs are found in the research of Speech recognition. Especially, Speech segment is regard as a pre-process or first condition of Speech recognition and application of Speech recognition to be researched frequently.In this paper, first an overview and analysis is given of existing typical algorithms for real-time Audio-Stream segment then a algorithm base on energy is proposed to Audio-Stream segment. The utilization of message-action finite state Automaton is also required to control different complex playback pattern.In this paper, simple and useful algorithms in all phases are given, The paper detail the design and the implement of the system. We develop a soft in visual C++6.0. We have did some experiments to tested that the soft could meet the requirement of real-time.