A comparison of two distinct approaches to the blind identification problem of underdetermined mixtures of cyclo-stationary signals with unknown cyclic frequencies

In this communication, we propose a new method to blindly identify the mixing matrix of a possibly underdetermined mixture of sources when input signals are cyclostationary with unknown cyclic frequencies. It relies upon a particular linear operator applied to the observations correlation matrix. Then, taking advantage of the properties of the above transformed matrix, a set of cyclic frequencies can be estimated, leading to estimations of the different columns of the mixing matrix. This latter is then estimated thanks to a classification procedure. Our approach is also compared with the PARAFAC decomposition one. Finally, computer simulations are provided in order to illustrate the effectiveness of the two proposed approaches.