Novel frequency-domain systems to recover temporally-white signals in multipath environments

In many digital communication applications, the transmitted signals (sources) must be recovered from the signals received at several antennas through radio channels that suffer multipath propagation. This problem can be solved in the frequency domain by using well-known BSS (blind source separation) algorithms for separating different instantaneous mixtures in all frequencies. Unfortunately, frequency-domain approaches suffer from the permutation/amplitude indeterminacies because the frequency domain sources can be extracted in a different order and with different amplitudes in some frequencies. We will show that there exits a statistically dependence between the sources in different frequencies. This dependence allows to solve the permutation/amplitude indeterminacies by adapting the coefficients of a MIMO (multiple input-multi output) system at each frequency bin in order to minimize the MSB (mean square error) between its outputs and the frequency-domain sources recovered at other frequencies. Using this idea, we will propose and compare two architectures for the frequency-domain separating system.