Research Of Algorithm Sound Source Localization Based On Microphone Arrays

Sound source location based on microphone arrays is a hotspot in acoustic signal processing filed in recently years. The microphone array has widely applied in sound/video frequency conference, speech recognition and enhancement. The source localization algorithm main research is the super-resolution direction estimation algorithms. A large number of high performance algorithm are produced. But the typical algorithms of MUSIC aims at narrowband system model, and it has a large amount of calculation. Therefore, the content of this article is about how to choose arrays structure and to reduce the computational complexity ISM algorithm of sound signal. Specific work is as follows:Firstly, the article discusses the study of array structure problems. Uniform linear array and uniform circular array which are often used, are analyzed and compared. The speech signal based on microphone array is near-field wideband mainly, so the article analyzes the impact of array element spacing, do the simulation according to several different distance of array element spacing. Then, it simulates near-field model of uniform linear array and uniform circular array, that is based on choosing proper array element spacing.Secondly, the paper researches the problems that complex computing is heavy, it is caused by the complex matrix direction vector in sound source location algorithm. The paper do the work that convertes the complex field to the real field of multiple signal classification (MUSIC) algorithm, through pre-processing method based on odd symmetrical array. Simulation analysis shows that direction vector will converted to real matrix as long as arrays meet to symmetry, regardless of the number of array is odd or even. The method is simpler, smaller computing amount and it can be applied to any symmetry of the array.Finally, because covariance matrix of the eigenvalue decomposition has a large amount of computing in ISM algorithm, and it is difficult to be used in the embedded system, Ordinarily power iteration algorithm, Inverse power iteration algorithm, Eigenvalue shift power iteration algorithm are analyzed and compared. Simulation analysis shows that eigenvalue shift power iteration algorithm can replace the eigenvalue decomposition, which reduces the computing complexity. This method has high accuracy compared with other two kinds of power iteration algorithm in the same situation. And the paper carries out experiment in Matlab simulation environment, that verifies the feasibility of the above-mentioned algorithm.