Distributed sensing for acoustic source localization in indoor reverberant environments

Abstract

The acoustic source localization problem relies on the estimate of the difference in time-delays of the signal received between a pair of acoustic sensors separated by a fixed distance. In the ideal case of a room with no reflections, one can identify the intersection of constant power level maps generated by multiple pairs of sensors to isolate the location of the source. The presence of reverberation however produces additional potential source locations that create uncertainty in the source location. In this research, impulse responses are simulated for a rectangular room using an image source model that incorporates frequency dependent absorption coefficients. These spatio-temporal impulse responses are applied in conjunction with room noise to simulate the signals recorded across a two-dimensional distribution of sensor pairs. The number of sensor-pairs and their spatial distribution that can optimally predict the source location in the presence of reverberant features is discussed.