New methods for signal separation

Optimal reception of a source in the presence of sources of interference emitting, entirely or in part, in the same frequency band, can be obtained by modeling the radiation diagram of an antenna network, thus protecting radio communications against jamming. In the absence of multiple paths, this gain matching consists of a linear combination of signals received by the antenna sensors. One of the most generally used methods for this modeling is to separate the source signals. The sources can be discriminated by making use of the fundamental differences between the source signals: decorrelation; different powers; different spectra; and statistical independence. The two first points require traditional source signal separation methods. They consist of orthogonalizing sensor signals. These are substantially sub-optimal. The last two points, that is to say spectral differences and statistical independence, have led the authors to develop original separation methods. There are very nearly optimal.<>