Research Of The Sound Source Localization Algorithm Based On The Microphone Array

Array signal processing is an important part of signal processing,it is widely applied in communication,radar,sonar and other civilian or military field.With the continuous development of array signal processing and people's growing demands,the microphone array technology arises at the historic moment.As one of the important branches of microphone array technology,the acoustic source location technology has attracted the scholars' attention and research.However,most of the current estimation techniques of DOA are based on the assumption of narrowband stationary signal.And it is not suitable for broadband short-time stationary speech signal.In view of this problem,this paper makes an in-depth study of the application of MUSIC algorithm in the acoustic source positioning system from the aspect of reducing algorithm computation,and verifies the algorithm through simulation experiment.The main contents are as follows.First,the speech source characteristics are analyzed,according to the different signal bandwidth,this paper introduces the signal processing model used in narrowband signal and broadband signal.And models of near field and far field are researched on the basis of uniform linear array.Second,due to the super resolution characteristics of MUSIC algorithm,in this paper,the MUSIC algorithm and its derivatives for root MUSIC algorithm,PM algorithm and MSCS algorithm is studied,the effectiveness of these algorithms is verified by experiment data and results.At the same time,from two aspects of computing complexity and localization accuracy,these algorithms are simulated for comparison,comparing the performance of them.Third,because of the broadband and short-time stationarity characteristics of speech signals,before processing the data with the traditional narrow-band DOA estimation algorithm,we should preprocess the collected speech signal in several methods,including window frame,speech endpoint detection,and the time-frequency transform.Then,the traditional sound source localization MUSIC algorithm is described.At the same time,for the shortcomings of large computational complexity,through the screening of frequency points and the optimization of the process of solving the covariance matrix to improve the algorithm,the experiment proves that this algorithm not only reduces the computational complexity,but it also has a good estimation performance under low SNR conditions.Fourth,a sound source localization algorithm based on angle pre-estimation is proposed,this method combines time delay estimation algorithm and improved algorithm.First,the estimation of the source azimuth is obtained through the generalized mutual correlation delay estimation algorithm,and a disturbance is added to obtain the search space of space spectrum.Next,a spatial spectral expression is obtained by using the improved sound source localization MUSIC algorithm.At last,searching the space of the space spectrum previously obtained and obtain the estimation of DOA.Simulation results show that the algorithm further reduces the amount of computation.