Abstract
A method for multiple sound source localization is presented. It is based on the properties of the onsets of sound as the primary cue: (1) most of the onsets will be contributed by only one source. (2) the onsets will be contributed only by the direct sound. Histograms of temporal disparities for several frequency components and its integrated one are calculated to determine the correct temporal disparities. Experiments were carried out in an anechoic chamber and an echoic chamber. The recorded speeches were used as two sound sources. We could localize the two concurrent sound sourcess with reasonable accuracy.
