Design of a real-time music-based sound local-izer for multiple sources in real environments

초록

In this paper a framework for localization of multiple sound sources in the 3D-space based on the MUSIC algorithm was implemented and evaluated to achieve the goal of improving speech communication in real noisy environments. The effects of ambient noise on performance of MUSIC algorithm for sound source localization in real-time performances were evaluated. A robust framework and methodology for real-time processing is proposed for determination of the multiple sound sources. The proposed framework achieved 3 degree localization accuracy in terms of azimuth and evaluation angles. This method also estimates the number of active sound sources and histogram of DOA estimates and showed that the results are consistent and repeata-ble with low standard deviation. The measurements were carried out in a real environments un-der different reverberant and background noise conditions to validate the practicality of the al-gorithmic chain and finally, the results were obtained and presented to demonstrate the efficien-cy of the proposed MUSIC based localization technique.