Abstract
This thesis examines three categories of acoustic source localization algorithms based on time delay estimation (TDE) models. TDE can be further classified into two approaches, time-of-arrival (TOA) and time-difference-of-arrival (TDOA). The first category of localization algorithms is the direct solutions of TOA and TDOA, trilateration and One-Step-Least-Square (OSLS) algorithm. While both algorithms have low complexity, they are sensitive to noise. The second category of localization algorithms are iteration solutions: Newton Method, Seidel iteration and Simulated Annealing (SA). Iteration solutions solve the acoustic source location problem in a recursive way and this can avoid dealing with inverse problem. Parameter estimation methods are the third category of localization algorithms and are separated into two part, the classical approaches and the Bayesian approaches. There are Spherical interpolation (SI) and Expectation Maximization (EM) based on the classical approaches, maximum likelihood estimator (MLE). Extended Kalman filter (EKF), iterated extended Kalman filter (IEKF) and Particle filter (PF) are acquaint as the Bayesian approaches. The EKF is suitable for real-time demonstration due to the high-accuracy and fast-computation. This thesis demonstrates two TDE experiments, ultrasonic gesture control and sound source tracking microphone array.