数字信号(国外英语资料).doc

数字信号 An overview of speech recognition Speech recognition is the technology that tries to make machines understand human speech. The function of speech recognition is to convert speech into equivalent written information, which is to make the computer understand human speech. As a cross subject, voice recognition and voice as the research object, the speech signal processing is an important research direction, is a branch of pattern recognition, involves computer, signal processing, physiology, linguistics, nervous psychology, artificial intelligence, and many other areas, even relates to peoples body language (for example, when people speak of facial expressions, gestures and other actions can help each other understand), its ultimate objective is to people and machines for the natural language communication Study history and the status quo of foreign studies The research on speech recognition dates back to the 1950s. In 1952 ATT bell LABS Audry system was the first voice recognition system to recognize ten English digits. Emerged in the late 1960 s, 70 s several basic thought of speech recognition is one of the important achievement in signal linear predictive coding (LPC) technology and dynamic time neat (DTW) technology, can effectively solve the long range of speech signal feature extraction and phonetic matching problem; At the same time, the vector quantization (VQ) and implicit markov model (HMM) theory are also presented. The study of speech recognition in the 1980s was further advanced: the HMM model and the successful application of the artificial neural network (ANN) in speech recognition. In the 1990s, with the rapid development of computer technology and telecommunication applications, it was urgent to demand speech recognition system from laboratory to practical. The most representative is IBMs ViaVoice and Dragon systems. These systems have a speakers ability to adapt, and new users dont need to train all the words to improve their recognition rate.