Design And Implementation Of Microphone Array Sound Source Location System

In recent years,with the rapidly increase of the number of cars in the city,the pressure of the urban traffic system has increased,and the problem of traffic noise pollution such as car whistle has also been brought.These problems have seriously affected the normal production and life of residents.At present,the most common way to identify and locate the car whistle noise is through manual judgment,which is inefficient and wastes human resources.Therefore,the method of locating the car whistle by microphone array is proposed and used in daily life.Through the microphone array,the sound data and spatial location of the sound source in the space can be obtained.The microphone array system for sound source location can use this information to calculate the actual position of the target sound source through the sound source location algorithm.In this thesis,the main work is to receive the audio signal in the road through the microphone array designed by ourselves,identify whether there is a whistle sound,and then calculate the specific direction and actual position of the vehicle.Firstly,this thesis introduces the structure,principle and method of car whistle recognition system based on convolution neural network.The energy-zero-product is used to judge whether the signal received by microphone contains whistle signal or similar high energy signal,and the acoustic characteristic parameters,MFCC+ GFCC,are extracted and combined with their dynamic characteristics.Use these parameters as the input layer of convolutional neural network,they are substituted into convolutional neural network model for training.Compare with the recognition results of the traditional BP neural network model through simulation experiments,it is show that the convolution neural network with mixed characteristic parameters has the superiority in the recognition of car whistle signal.Then this thesis introduced the optimization design method of planar microphone array based on genetic algorithm.The beamforming algorithm of far-field delay-sum is used to analyze the maximum beam width,maximum sidelobe gain and other parameters of microphone array.They are used as optimization functions to design the corresponding genetic algorithm in this thesis.The uniform-random combined two-dimensional microphone array optimized by this method has a narrower maximum beam width and a lower maximum sidelobe gain.Compared with the existing microphone array on the market,it has more excellent performance.Next,this thesis builds a microphone array signal acquisition system by using MEMS microphones.According to the structure of microphone array optimized by genetic algorithm,44 element plane MEMS microphone array is made in this thesis.Zynq 7z035 is used as the main control module to realize the synchronous signal sampling of 44 microphones.The data including the whistle signal is transmitted to the upper computer at high speed through USB2.0 interface.At the same time,the upper computer software of signal receiving is made,which can store and decode the received audio signal.In the end,this thesis,using car whistle do some experiment is carried out,and simulate the car whistlem,use the sound source localization algorithm to process the data collected by array.The microphone array sound source location system designed in this thesis can accurately locate the position of the target sound source by using MUSIC and other sound source location algorithm.