Circumferential Multiple Source Separation Using a Small Microphone Array and 2-D Spectra
Senta Ariizumi, Teruki Toya, Kanji Watanabe, Shuichi Sakamoto, Kenji Ozawa · 2024
This study investigates sound source separation using a compact microphone array to address the challenges of closely located sources. The proposed method aims to solve simultaneous equations related to the two-dimensional (2-D) spectrum by treating microphone outputs as spatio-temporal sound pressure distribution images and analyzing their 2-D spectra. Experimental evaluations under known source directions show promising results for the separation of multiple sound sources. The results indicate stable performance with an average signal-to-distortion ratio of approximately 20 dB at narrower angular separations.