A new consistency measure for localization of sound sources in the presence of reverberation

Abstract

Sound source localization is an important task in many practical applications. Among existing methods, the Linear Intersection algorithm is a popular technique and performs well in noisy conditions without requiring high computational costs. Unfortunately the performance of Linear Intersection is seriously hampered by the presence of reverberation, which is quite common in closed environments. In order to overcome this limitation, the Optimal Line Selection technique was proposed in the past. In this paper some important modifications of this technique are presented and described. Results on simulated and real data confirm the quality of the proposed solution.