Three-Dimensioinal sound source localization using Inter-Channel time difference trajectory

Abstract

This research addresses a three-dimensional sound source localization (3-D SSL) using the inter-channel time difference (ICTD) trajectory which is a new localization cue proposed in the research for mobile applications. To be an efficient 3-D SSL, as an attempt to decouple 3-D SSL into two disjoint azimuth and elevation angle estimations, the (source) directionand (microphone) position-dependent ICTD trajectory was realized by 2-channel rotating array installed on a spherical platform. The mathematical formula of the realized ICTD trajectory shows that the trajectory’s mean and phase shift are strongly related to the azimuth and elevation angles respectively in a form of one-to-one relation. Besides, the cone of confusion problem that has not been solved by the previous SSL methods using two microphones can be treated well. In addition, the probabilistic approach for 3-D SSL was investigated on the basis of the Bayesian inference. As a result, the inference-based SSL method was sufficiently well-adapted for the 3-D SSL method using ICTD trajectory: the prior probability of a source direction and a likelihood function of measured ICTD trajectory given possible source directions. In addition, the density-based spatial clustering of applications with noises was modified to cluster the measured ICTDs and reject the noisy ICTDs. By the simulation results, it was found that the SSL performance is strongly dependent on the azimuth angle only and it was quite reasonably interpretable by the trajectory’s amplitude: the left-sided sources can be estimated better than the right-sided sources considering the configuration of the 2-channel rotating array. In the realized system, the 2-channel rotating microphone array is composed of the wireless microphone system, the ultrasonic motor, and the encoder. To validate the proposed method, two experiments were conducted in the room environment: single source case using a Gaussian white noise and a male voice source; two source ...