Sound Source Localization Based on PSVM algorithm

Abstract

Sound source localization is one of the major audiovisual functions of intelligent robots. With the development of computer technology, high-quality sound signal acquisition technology has been widely used in many fields such as microphone array sound source localization. Support Vector Machine (SVM) is a kind of machine learning based on statistical learning theory and structural risk minimization principle. Many parameters will affect the performance of SVM. By changing parameters, the ability of anti-noise can be improved. However it comes at the expense of computation and speed. Proximal support vector machine (PSVM) is an improvement on classical SVM. It has smaller amount of calculation and faster speed. In this paper, by extracting the characteristics of generalized cross-correlation function of sound source signal and using PSVM to locate the sound source, the sound source localization accuracy is better in the reverberation and noise environment, and it has good robust performance.