Design About Application Of Speech Recognition In Vehicles

The Speech Recognition is a kind of technology, which makes the machine change the speech signals into homologous text or order by recognition and comprehension. And as s a cross-subject, it is becoming to the key-technology of human-computer interface in information technology. The combination of speech recognition and speech synthesis make it is possible that people can operate the computer through the speech instruction without the keyboard. Thus, security and comfortableness for the automobilist will be enhanced if the dividers control the. equipments of vehicle through the speech instruction.At present, according to the difference of hardware, the research of Speech Recognition System includes: the Speech Recognition System based on PC and the embedded Speech Recognition System based on special chip. The research about the embedded Speech Recognition System in the text is based on TMS320VC5402 DSP(Digital Signal Processor). The text proposed one kind of speech recognition system application settlement, and introduced its hardware design and the software realization, and pointed out the prospect of application on the automobile.Firstly, this paper introduced the present situation about the development processing of speech signal research in the domestic and foreign countries, and point out research background and its significance. Further more, according to the speech recognition system constitution model, the paper introduced fundamental theory about the speech recognition, including the speech signal pretreatment, the speech endpoint examination and the Feature Parameter Extraction. And it analyzed the LPCC (linear-prediction cepstrum coefficient) and the Mel Frequency Cepstrum Coefficient with emphasis, introduced the Vector Quantification theory, and elaborates the Vector Quantification basic principle and best code book design in detail.Secondly, brief explanation to TI Corporation's DSP, the article proposed the corresponding recognition system plan which referrer to the mature theory foundation existing. The hardware of the whole embedded system mainly includes the following units: the power module, A/D conversion module, expansion memory module, control and communication module and so on.Moreover, in the software algorithm, the paper adopted the commonly used endpoint detection method based on SAE-SAZ(short-time average energy and short-time average zero-crossing rate), and take the Mel cepstrum coefficient as the speech feature parameter, moreover, based on the requirement of speaker-independent embedded system, in order to reduce memory capacity and calculation,, the feature parameter is compressed by vector quantity. The discrete hidden markov model (DHMM) is adopted, Baum-Welth reestimation algorithm, forward-backward procedure and viterbi algorithm is utilized to train and recognize the speech signal. Then the entire algorithm may achieve the recognition duty. The whole algorithms are emulated in PC by MATLAB.Finally it introduced software design in DSP, summarized the full text, and pointed out the future work and the direction to be researched.