Detection of acoustic signal source by use of cross spectra between every pair of sensors distributed in space

This paper describes the method to find out the outbreak of a sound source and to estimate the position of the source in noisy environment using distributed microphones. A source point is first assumed and the cross spectrum between the outputs of every pair of microphones is computed after compensating the spectral change by transmission path from an assumed source point to the microphones. Then the cross spectra are averaged. Scanning the assumed point, the averaged cross spectrum is computed for every assumed point. And the source point is estimated as the point of maximum magnitude of the averaged cross spectrum. The outbreak of a sound source is detected by the change in the averaged cross spectra. It is a practical problem that a number of computation of FFT is necessary in the proposed method, and this problem is solved by developing a high speed pipeline processor. The effectiveness of the method is confirmed by the computer simulation and by the experiments.