Biological versus Subspace Methods in Sound Localization

Abstract

Sound localization is determining the location of sound sources usingthe measurements of the signals received by an array ofsensors. Humans and animals possess the natural ability of localizingsound. Researchers have tried to model nature's way of solvingthis problem and have come up with different methods based on variousneuro-physiological studies. Such methods arecalled biological methods. On the other hand, there is another community ofresearchers who has looked at this problem from pure signalprocessing point of view. Among the more popular methods for solvingthis problem using signal processing techniques are the subspacemethods. In this thesis, a comparative study is done betweenbiological methods and subspace methods. Further, an attempt hasbeen made to incorporate the notion of head-related transfer functionin the modeling of subspace methods. The implementationof a biological localization algorithm on a DSP board is also presented.