Design And Development Of Vehicular Speech Recognition System

With the rapid development of automotive technology, more and more control technology is applied to car to improve the safety, ride-comfort and entertainment. During the recent years, speech recognition technology has made remarkable progress, and speech products begin from the laboratory to the market. Meanwhile, this technology is paid more attention to apply to automotive system to implement the intelligent control to the partial electronic & electric equipments. This paper is about the design of car speech recognition system based on the core-controller SPCE061A produced by LingYang Corporation with the speech recognition method of DTW, and this system can fully control the automotive electronic&electric equipments intelligently by the simple speech orders of driver.At first, the present primary speech recognition methods are generally introduced, and the form and principle of typical speech recognition system are systematically illuminated in the paper. In order to obtain the abundant and effective information in the complicated speech signal, speech signal processing technology is particularly introduced. Firstly, the pre-processing of speech has been expatiated including pre-emphasis processing, end-point detection processing using the parameters of short-time average amplitude and zero-crossing rate, and the speech framing process; secondly, two kinds of speech feature parameters extraction process are discussed; finally, simulation process of the scheduled voice is done with the help of Matlab software to lay the foundation for the realization of programme of speech recognition.According to the feature of speech recognition system which is designed in the paper, DTW speech recognition method is determined as the recognition method of this system compared to other speech recognition methods, and the principle of DTW arithmetic is generally discussed, and the searching process of matching path is particularly analysed, and the estimation rule of speech recognition result is given. According to the shortcoming of DTW'vast calculation and data storage, a new modified DTW arithmetic using subsection warping and searching the definite parallelogram domain is carried out to implement the model matching to adapt to the demand of system hardware resource.The scheme of hardware design is achieved independently in the paper. Under the guidance of this scheme, system hardware selection and circuit principle charts are accomplished including the selection of core-controller SPCE061A, the design of stable voltage power system using integrated stable voltage chip, the design of the circuit of audio input and output module, the design of LED display module, the design of drive control module, the design of keyboard input and communication module. And experiment testing is fulfilled to check the correctness of design. Finally, the hardware platform of speech recognition system is builded up through connecting each function module.On the basis of integrated development entironment IDE1.8.4 supplied by SPCE061A, the programme of speech recognition system is compiled with C language and assembly language. Firstly, the general flow chart of system programme is determined, and the drive programme and arithmetic programme are respectively finished under the guidance of this chart. The drive programme includes the initialization of I/O, keyboard setting, ROM and Flash, and the arithmetic programme includes speech training programme and speech recognition programme.At last, in order to inspect the recognition effect of car speech recognition system, two experimentation schemes are designed to carry out the recognition detection. Firstly, fifty-time recognition detections have been operated to each two -word speech order and the results have been taken note of. Then the experiments have been repeatedly done under the condition of changed background noise. Continuely, recognition detections have also been operated to each four-word speech order. Finally, the recognition rate has been calculated based on the experiment results. From the calculation result, the comprehensive recognition rate of car speech recognition system is more than 83 percent, geting a considerable recognition effect.