Design Of Chinese Speech Recognition System In Vehicle Interior Noise Environment

Language is the most important communication tool，also the most primitive and simpleway in human society. In the history of human society, people always want to find someways to make the communication between themselves and nature come true and make thedeeper exploration of nature. Under the push of the desire, machine was invented and widelyused in the production and society activities as a new kind of tool for the communicationbetween human and nature. The use of machine greatly accelerated the progress of humansociety. The relationship between people and machine becomes more and more subtle.With the improvement of the producing and life, people are no longer satisfied withmonotonous and boring control of machine. People are getting more interested in theman-machine conversation. The development of electronic system engineering theory andtechnology make it possible. And language was made into the best tool between the two. Tomake the machine can have conversations with us we should make sure that machine canunderstand the words of people. This is what scientists have been working on for many years:teach the machine to understand what people say. In technology it is to tell machine how torecognize and learn the words and then how to turn the words into some information likecontrol instructions or text information. The theory originated in1950s and it has beengreatly improved after more than half a century of development. Some mature productionshave been on the market in mobile-phone and computer field.As one of the most important machines have the closest relation with human, cars havebeen focused a lot and been asked more on entertainment and the comfort aspects, so thatsome new technologies such as the speech technology is gradually used in vehicular field asa new advanced thing to satisfy people’s demands. In this paper, a Chinese speechrecognition system in a noisy car environment with the core processor TMS320C5509A ofTI is proposed. The core algorithm of the system is the theory of dynamic time warping(DTW). And LMS (Least mean square) filter module is designed for the system according to the interior noise to filter the noise to get a stable recognition effect and a better recognitionrate. A series of right-sizing auxiliary device control instructions are designed for recognition.After recognized, the results are sent to CPU for judging and then corresponding controlinstructions will be generated so that the auxiliary equipments can response to user’sdemands. It not only improves driving pleasure but also provides facilities for handicapped.Firstly the history, basic theory and speech signal processing technology are introducedin the paper. And the improved DTW algorithm is also discussed. Then the hardware withthe core processor DSP and extended with an audio system and storage system and etc isproposed. Next was the detailed presentation about the design of software system. Westudied the characteristic of automotive interior noise and designed filters with the help ofMATLAB. Also the simulation of the recognition in MATLAB was carried on. In the end wefinished the programming in C language in the CCS development environment. After onlinesimulation, a result was got and the code was burned into FLASH to make the bootloadercan read the boot table and then transfer code from external FLASH into internal memoryfollowing power-up.