| With rising ownership of vehicles across the country over the years,the urban road traffic noise problem becomes more and more serious.The study has found that noise pollution can have a great impact on people’s living and working,as well as many health effects.As in the traffic noise,the horn occupies the vast majority,if the illegal vehicle horn phenomenon can be effectively curbed,will greatly reduce the noise pollution.Vehicle honking identification and sound location system is an intelligent capture system for noise,honking and other sound sources identification in the high-tech era.This paper mainly studies the honking recognition and sound location algorithm of vehicle honking recognition and sound location system,and builds an experimental platform to verify the system.The main work of this paper is as follows:(1)In the part of whistle recognition,puts forward a horns recognition method based on support vector machine(SVM).Car honking and urban traffic noise are analyzed.And starting from the harmonic features of the distribution characteristics of car horns,extract the Melfrequency cepstral coefficient(MFCC)feature as a feature vector,using support vector machine(SVM)as a car horn and traffic noise of two classifiers.At the same time,the paper analyzes the traditional MFCC features of Mel filter number and dimension of the impact on the recognition effect.It is found that the honking recognition effect can be improved by increasing the feature dimension.(2)The sound source localization algorithm based on time difference of arrival(TDOA)is improved.The microphone array and sound source signal model are introduced.Analyzed the basic principle of generalized cross-correlation and improve the cross-power spectrum phase method(MCSP)were compared.It is found that MCSP has better stability and accuracy of delay estimation under low signal-to-noise ratio(SNR)condition.Then,common fractional time delay estimation method is studied,including the correlation peak interpolation method and parabolic fitting.The spherical interpolation method combined with the design of the fourelement microphone array structure is used as the location estimation method in the honking scene.A distributed microphone array method was proposed to solve the problem of large distance estimation error of spherical interpolation method in the far field.At the same time,analyzes the microphone position error and the position error between distributed array affect positioning estimation.(3)In the realization part of the system,an image marking method of honking vehicle is proposed.By means of camera calibration,the three-dimensional coordinates of the honking sound source can be converted to the camera coordinate system,and then image interpolation can be used to realize the image marking of honking vehicles.Then the experimental platform of vehicle honking recognition and positioning system is built,and the related experiments of honking source identification and positioning and image labeling are carried out.The experimental results show that the system designed in this paper can complete the functions of recognition,sound location and image marking.Finally,the errors in the experiment are analyzed. |
