Research On Traffic Volume Detection Technology Based On Driving Noise

With the rapid development of intelligent transportation,more and more attention has been paid to the research of traffic volume detection using audio detector to obtain driving noise signal.At present,traffic volume detection based on driving noise is usually carried out in quiet and low flow road environment,rarely involving continuous traffic flow and overlapping depot.Therefore,in the follow-up research,it is necessary to explore the methods to improve the signal-to-noise ratio,find the appropriate characteristic parameters,and study how to improve the detection accuracy of continuous traffic flow and overlapping depot.To solve these problems,the main contents of this paper are as follows:(1)Based on the analysis of the components and influencing factors of the driving noise signal,an improved spectral subtraction method based on multi window spectrum estimation is proposed to reduce the noise of the collected signal.Firstly,the noise signal of the traffic is preprocessed,and the loss of high frequency component is compensated by the pre weighting;The nonstationary signal is converted into stationary signal by using frame adding window;Then the noise signal of the pre-processing is reduced.Finally,the experimental results show that the improved multi window spectral subtraction method has more obvious effect in complex environment.The research of this part is to make a preliminary preparation for feature extraction of driving noise signal.(2)Based on the analysis of different features in three dimensions of time domain,frequency domain and transform domain of driving noise signal,a method of extracting principal component feature fusion is proposed.In the analysis and research of the short-term energy,short time zero crossing rate,sound spectrum and Mel cepstrum coefficient of the driving noise signal,it is found that the single feature has some defects such as incomplete representation information and data not obvious;The proposed extraction method based on the feature fusion of principal component,which combines multiple feature quantities through PCA,then integrates them into a new feature quantity.In the case of almost no increase of the computation amount,the useful information of signal extraction is improved.(3)Through the research on the characteristic quantity of traffic noise signal,an innovative traffic volume detection method based on triangle wave analysis of traffic noise is proposed.Firstly,the extracted short-term energy and Mel cepstrum coefficients are fused to form new features by principal component feature fusion method.Then,the peak value is found in the time series of the feature curve.Then,the overlapped depot signals are separated by triangular wave algorithm.Finally,the traffic flow is detected by calculating the number of triangular waves.The experimental results show that the accuracy of the traffic volume detection method based on triangle wave analysis reaches 93.99%,which can meet the actual demand of traffic volume detection.