Comparison second order based blind signal separation with classical adaptive interference cancellation methods in the case of ill-conditioned statistics

In the last years blind source separation methods increasingly frequently use in digital signal processing. Their advantage is that we haven't to know any additional information about the source signals. The BSS method uses two fundamental presumptions. The first one is that the observation signals are linearly dependent on source signals. The second presumption is that the source signals must be independent from each other. The possibility of source separation using just observe signals let to decrease systematical error which correlate with the wrong data of antenna array. The purpose of this paper is comparison the BSS method with another one and efficiency of modification the BSS method with the Tikhonov regularization. The MVDR and Timegate methods were chosen for the comparison with BSS method. The experiment was run in the Matlab. Two sinusoidal mutually spaced signals fall into uniform linear array. The maximum signal-noise ratio was chosen as the criterion. The experiment shows that BSS method better separate the signals that the other ones. In the second part of the paper BSS method was analyzed in the case of ill-conditioned statistics. This situation is possible when the number of antenna elements is larger than the number of source signals. An experiment was run in the Matlab where the rate of off-diagonal elements of the statistics was calculated after the diagonalization. The experiment shows that the Tikhonov regularization essentially decreases the summar off-diagonal elements rate and improves source separation in case of ill-conditioned statistics.