Blind identification of colored signals distorted by FIR channels

This paper presents a new approach for blind identification of multiple colored stationary/nonstationary signals distorted by unknown FIR channels. The key idea of this approach is to use a bank of decorrelators to transform a multiple-input and multiple-output system (driven by the desired signals) into a bank of single-input and single-output systems. This approach is referred to as BID, i.e., blind identification via decorrelation. The BID approach can uniquely (up to a permutation and scaling) identify the signals if (a) the signals are mutually uncorrelated and of distinct power spectra, and (b) each column of the system function is a coprime polynomial vector. No other method known to date can uniquely identify the signals under the above condition. The BID approach achieves the optimal identifiability potential recently predicted by Hua and Tugnait.