Implementation of TDOA-based acoustic source localization in Java

Abstract

The aim of this project is to develop a Java program for acoustic source localization (ASL) technology. In this project, a source localization method was implemented using Java and this implementation was further used to create an Android application. The implemented localization method is based on the time-difference-of-arrival (TDOA) approach. Generalized cross-correlation (GCC) and Levenberg-Marquardt algorithm (LMA) are well-known techniques used for achieving TDOA-based ASL. We estimate TDOA by processing signal data using GCC with phase transform (GCC-PHAT) technique, and subsequently achieve ASL using LMA-based optimization. We analyze the mathematical features of the GCC and LMA techniques. We implement the two techniques using Java for software-based ASL as Java is famous for its system independency. Upon completion of the program development, we use the program to simulate ASL. Based on the result, we evaluate the behavior of the program and analyze the performance of the implemented ASL approach.