Blind 3D sound source direction using stereo microphones based on time-delay estimation and polar-pattern histogram

Abstract

In this research, a new 3 dimension sound source direction algorithm was developed for computing the azimuth and elevation direction of the source under a limited space between two microphones in noise free environment. The proposed azimuth-estimation method is based on the time — delay estimation using maximum likelihood estimation by forming a histogram of power weighted spectrum corresponding to attenuation and time-delay index. For estimating elevation angle, the proposed method requires a sole recorded signal which selected the higher sound pressure level between the two recorded signals. The elevation-angle estimation is computed by phase delay of the windowed autocorrelated signal. The windowed autocorrelation process of the selected signal benefits for reducing the influence of the artifacts. A polar-pattern histogram is employed for identify the source position associated to the microphone i.e. above or below. The proposed method has been demonstrated by real-audio signals. Given by a narrow space, the proposed azimuth-estimation method renders a significant better directional performance improvement over the benchmarked method.