Mixed-signal VLSI microsystem for acoustic source separation

We present an architecture of a microsystem designed for the task of blind source separation of acoustic sources that impinge on the miniature microphone array. The proposed architecture implements the source separation algorithm in a unique framework that combines wavefront sensing, subband signal processing and independent component analysis. The spatial gradients of the acoustic wavefront are computed in the continuous-time domain and decomposed in the 16 subbands. In each band, we perform static independent component analysis. We demonstrate that with the proposed system architecture 13 dB separation performance is achieved in moderate reverberant environment.