Blind source separation of 3-D located many speech signals

This paper presents a prototype system for blind source separation (BSS) of many speech signals and describes the techniques used in the system. Our system uses 8 microphones located at the vertexes of a 4 cm times 4 cm times 4 cm cube and has the ability to separate signals distributed in three-dimensional space. The mixed signals observed by the microphone array are processed by independent component analysis (ICA) in the frequency domain and separated into a given number of signals (up to 8). We carried out experiments in an ordinary office and obtained more than 20 dB of SIR improvement