Research Of Sound Source Localization Based On Microphone Array

Speech acquisition is an important part of speech signal processing.Sound source localization technology based on microphone array beamforming is an effective method to improve speech acquisition performance.It has become a research hotspot in recent years.Due to the movement of the sound source position,the reflection and diffraction of the sound and the change of the external background noise,it is difficult to estimate the sound source location and the number of sources.Based on the in-depth study of the conventional MUSIC algorithm for microphone array beamforming,this thesis proposes an improved MP-MUSIC algorithm,which solves the problem of sound source localization accuracy and source number estimation.This thesis mainly completes the following work:1)Analysis of current development status,background and existing problems of sound source localization technology at home and abroad.The basic principles and methods of sound source localization are introduced.Different sources number estimation methods are compared and simulated.The results show that the method of estimating the number of sources of the constant false alarms has better performance.2)Set up the mathematical model of the microphone array,modeling and simulation of the linear microphone array and the circular microphone array in MATLAB respectively.The simulation results show that the circular array has a wider range and larger space than the linear array in the positioning range.3)An improved MP-MUSIC algorithm is proposed.Using the basic idea of match tracking,the signal is represented by linear sums of multiple column vectors,and a new covariance matrix is formed.The eigenvalues are then obtained by eigen decomposition.Corresponding eigenvectors,based on the estimated number of sources,can separate the noise subspace from the eigenvectors,use the new noise subspace to construct the spatial spectrum,and seek the peak value to obtain the estimated value of the direction of arrival.Through MATLAB simulation,compared with the traditional MUSIC algorithm experimental results show that the improved algorithm in the positioning accuracy has improved,and with the reduction of SNR,its positioning performance is better than the traditional algorithm,in the simulation of real voice environment,The improved algorithm can use beamforming to track sound sources and suppress noise.4)The design of the hardware structure includes the design of the hardware block diagram,the selection of the processor,and the design of the circuit structure.The hardware uses DSP6000 as the main processor,and the main control circuit,digital-analog conversion circuit and timing control circuit are designed.Using Microsemi platform software to test the hardware and improved MP-MUSIC algorithm,the experimental results show that the improved MP-MUSIC algorithm's positioning performance is better than the traditional algorithm.