Ohnishi Lab. |
Computer Audition Research Theme |
Multiple Sound Source Localization and Separation from their Mixture |
This paper presents a system for sound source localization and separation inspired by the auditory mechanisms of biological systems. The system consists of three omnidirectional microphones, banks of band-pass filters, and a personal computer with a digital signal processor. Sound sources are localized using arrival temporal disparities. The zero-crossing method is used to quickly detect the arrival temporal disparities at signal onset to cope with echo . Arrival temporal disparity histograms remove ambiguity and improve accuracy. The Histogram Mapping Method is introduced to localize multiple sound sources. The system separates multiple sound sources by obtaining their direction information from the localization process. Localization and separation experiments were conduced in an anechoic chamber and an empty room. Two sound sources, human voice, were used. Directions of the sound sources were localized within a few degrees. Sound sources were separated by 25 dB attention. |