Sound-source localization system based on neural network for mobile robots

Abstract

In this paper we described a sound-source localization (SSL) system which can be applied to mobile robot and automatic control systems. A novel approach of using artificial neural network was proposed to obtain the horizontal direction angle (azimuth) of the sound source. According to humanoid characteristic only two microphones, which were attached symmetrically on both sides of the robot as its two ears, were utilized and tested. Sound wave signals were received from both microphones and analyzed directly by neural network. Two sets of training data were collected and used to train neural network, according to which, different performances of the SSL system were verified and compared. The strong recognizing and calculating ability of neural network made the system work effectively and accurately.