S. Benton, A. Spanias, K. Tu, H. Thornburg, G. Qian, and T. Rikakis (USA)
real-time sound localization, particle filters, acoustic beamforming, microphone arrays.
: This paper addresses the problem of localizing a speech source in a reverberant environment. Simulations and real-time experiments are conducted and several methods are examined including the traditional steered-response-power (SRP) and the delay-and-sum beamformer (DSB). An algorithm that uses a particle filter in conjunction with an adaptive beamformer is developed, analyzed, and tested in a reverberant environment. The use of the particle filter provided improved results, relative to traditional DSB localization methods, in offline simulations and real-time experiments. The rest of the paper is organized as follows. Section II defines the signal model and provides formulations of the DSB and the SRP-PHAT methods. Section III defines the general particle filter framework along with some implementation details. Section IV describes several experiments and presents comparative results. Section V gives the conclusions of this study. 2. SRP SOURCE LOCALIZATION A. Signal Model For a source signal s(t) received at an array of M microphones, the collection of microphone signals xi(t);i=1,2,..,M-1, can be written as,
Important Links:
Go Back
