TDOA and beamforming based acoustic source localization and enhancement using a linear microphone array

Abstract

Acoustic source localization (ASL) and enhancement are sub-disciplines of acoustic signal processing. ASL is an important technique to localize the position or the direction of the acoustic source, while acoustic source enhancement (ASE) is aimed at extracting a desired sound from a given direction. This project studies the source localization and enhancement for a single stationary acoustic source using a linear microphone array, and involves two types of methods-time-differences-of-arrival (TDOA) and beamforming. The TDOA method is illustrated by introducing generalized cross-correlation (GCC) with phase transform and Levenberg-Marquardt (LM) algorithm. The beamforming method in this project includes the delay-and-sum and steered response power (SRP) beamformers for localization, and the delay-and-sum and minimum-variance distortionless response (MVDR) beamformers for enhancement. When presenting these four ASL and ASE approaches, this thesis first demonstrates the theory of each approach before describing the intended application. In this work, MatLab is the main tool used for simulation. The GCC with phase transform along with Levenberg-Marquardt algorithm (GCC-LM) and delay-and-sum beamforming are simulated in a virtual room set up by MatLab, while the SRP beamforrning and MVDR beamforming are applied in real environments using the MatLab platform. When introducing different ASL and ASE techniques, this thesis also makes comparison among them to help make decision on which ASL or ASE technique is the most suitable according to the requirements of a particular application.