Identification aveugle de filtres ARMA et applications

Abstract
The blind identification of linear systems is an important problem in signal processing (communication, seismic,..). The second order statistics (Autocorrelation function, or Power Spectral Density), are blind phase, they only permit the identification of minimum phase linear time invariant systems. During the past years there has been an increasing interest in utilizing the Higher Order Statistics (HOS), to solve the problem of non-minimum phase blind identification. Because the HOS keep information about the phase. To identify ARMA filters with unknowns orders, we use all the information given by the second order statistics, and complete it by searching to maximize the kurtosis. We apply this approach to field seismic data.