Small sound extraction using virtual counter sound source method in loud noise environment
Keiichi Zempo, Yukino Hirai, Takuto Sakai, Koichi Mizutani, Naoto Wakatsuki, Hiroshi Yoshinaga · 2015
The purpose of this research was to develop a system that can detect the voice of victims in noisy disaster sites in order to facilitate more rapid rescues and decrease casualties from disasters. In this paper, the improved Virtual Counter Sound Source (VCSS) method, which is based on the delayed-sum method, was proposed to separate the small sound source immediately. This proposed method reduces the number of iterations by assuming the view angle of the sound source. Through an experiment using a quiet human speaking voice and a loud car sound, it was confirmed that the iteration number decreased in the proposed method.