A Dezign Of Detection Device Of Vehicle Collision Bases On Sound Signals

Traffic injuries are increasingly becoming a threat to the safety of human life and a worldwide public hazard. With the dramatic increasing of the number of cars, traffic density to significantly increase the traffic congestion problems are worsening tensions, and traffic accidents are increasing year by year. When the accident happens report to the traffic police timely often can be effectively assistance, so that development of high real-time vehicle collision detection device to the timely completion of the alarm after the accident can play an important role on reducing accident mortality and disability rates.This stage, the accidents are mainly based on the speed detection device signals, video signals and some other detection methods, but these methods have their problems exist, such as low time performance, low accuracy, then result to the difficultist to obtain a timely and effective rescue after a traffic accident.Because traffic accidents are allways with great sound of the collision, while the collision frequency noise and other sounds not the same. Through the collection and analysis of acoustic signals around the vehicle can be a method to detect vehicle accident. The scene in real-time access to information and alarms, so in real time of ways than the traditional methods, and the success rate of recognition in the accident may be higher. Based on the above reasons, the completion of the following studies:(1) Analyze the acoustic signal pattern recognition system and the basic composition, and understand the vehicle collision based on acoustic signal detection theory in-depthly. And do research on a variety of signal preprocessing, feature extraction and classifier design method.(2) Design of an algorithm to classify collision of acoustic signals based on wavelet feature extraction. This algorithm of the acoustic signal was received 7 level wavelet decomposition characteristics of the 18 frequencies, and the use of impact acoustic signal from a lower correlation coefficient characteristics and energy extraction characteristics of additional features. In the classifier design, the use of LDA, SVM and OC-SVM were classified try to compare the classification obtained the optimal form.(3) Complete the hardware design and software design of the vehicle collision detection system based on acoustic signal. The system uses TI DSP hardware company TMS3205509 as a central processing unit and TLV320AIC23B as sound acquisition module chip, and GPS and GPRS modules can communicate with the DSP module and alarm when collision happens. complete of the system's software design base on the system's hardware design and achieve the above-mentioned algorithms.(4) Finally, do experimental test of the system. The experimetal content algorithm simulation in Matlab and test the system in real evironment. Experimental results show that the algorithm designed in this paper could give the test results in real time and has high accuracy, which can effectively separate the collision signals from non-collision signals. In the system test in real evironment, the vehicle collision detection device is demonstrated to have stable performance, it could meet the practical requirements, so that the device has a good prospect.