Blind source separation for frequency dependent channels

An algorithm is developed for the blind separation of combinations of independent signals traveling over frequency dependent channels. The algorithm relies on the simple observation that if signals are independent in one domain, then their corresponding components in a linear transformed domain are also independent. The algorithm lends itself to the computationally efficient FFT implementation, which is given in this work. In addition, permutation indeterminacy inherent in the frequency domain blind source separation (BSS) is resolved by using cross-statistics between the source frequency components. Due to gain indeterminacy, the recovered signals after IFFT are the linear filtered versions of the source signals. The signals are postprocessed by an equalizer to solve this problem. Simulation results are given, which confirm the effectiveness of the proposed technique