Acoustic source separation via particle velocity vector measurement
Kenbu Teramoto, M. D. Tawhidul Kahn · The Journal of the Acoustical Society of America · 2006
A novel blind source separation of a mixture of two or more voice signals has been proposed in the present paper. The separation system has been focused based on the spatio-temporal blind source separation algorithm. The proposed algorithm utilizes the linearity among the signals of sound pressure and the components of three-dimensional (x, y, and z directional) particle velocity, all of which are governed by the equation of motion. Principally, as the mechanism of blind source separation uses no a priori information about the parameters of convolution, filtering, as well as mixing of source signals, some simple assumptions such as the statistical independency of the linearly combined (mixed) observed signals containing zero mean as well as unit variance have been implied in the present separation algorithm. Therefore, the proposed method has successfully simplified the convoluted blind source separation problem into an instantaneous blind source separation problem over the spatio-temporal gradient spaces. An acoustic experiment with two female voices has been carried out to compare the simulated data as well. A Micro-Flown system has been adopted to evaluate the voice signals efficiently.