| As the economy growing of our country, people's living standards have markedly improved, so many people like to have cars of their own,the number of cars greatly increases,at the same time, the problem of the congestion on the highway, gradually plug of people's attention,so the highway vehicle flow statistics also started .At present, the main traffic statistical methods of observation are:artificial observation , shift observing, video, automatic counter . Accurate statistics of the vehicle flow can not only to avoid overcrowding, congestion, save people's valuable time effectivly, but also conducive to the maintenance of the road maintenance.In this paper, firstly, we do endpoint detection using the vehicle noise ,find out the start and end position of the vehicle noise,intercept the effective part of the voice, and then extracting parameter from AR spectrum analysis,senting the parameter to the trained BP Neural network for identification,finally,calculating out the number of each model of the vehicle.The whole system Realized in MATLAB platform,based on the perfect toolbox,integrated the each module of AR model spectral analysis, BP neural networks and statistical models.For the traditional statistical methods,the system of this paper can be simple and rapid realization of the functions of statistics,significantly reduce development time and costing,so have great potential for development.This system is suitable for cycling road and small flow of highway, can be more accurate to identify the basic models on the road.Because of limited time and knowledge, the interface and other aspects of the system remain to be further improved. |
