Sources lndentification : A Solution Based on the Cumulants

In a great number of situations (array processing, seismic sounding...), the signal received on an array of sensors is the sum of several contributions (sources). In general, it is realistic to assume that the sources emit independent stochastic signals. In this case, several methods using the second order statistic of the signal have been developed in order to detect the number of sources and to identify the sources. It is known that the source identification is then not feasible in the general case and some hypotheses (plane waves ...I must be added. When the signals emitted by the sources are not gaussian, the second order statistic does not contain a complete description of the statistical properties of the signals. We show that, in the non gaussian case, it is possible to identify the sources using the third and fourth order cumulants. We illustrate the possibilities of this method in simulated situations.