Multichannel blind signal deconvolution using high order statistics
Éric Moreau, N. Thirion · 2002
The problem of multichannel blind signal deconvolution is considered. We show that input signals can be restored (or separated) using only the condition that they are statistically independent. Two main necessary and sufficient conditions involving high order cumulants are given and proved. Hence, a class of criteria for multichannel signal deconvolution are obtained. Self adaptive gradient based algorithms are derived in order to optimize the proposed criteria and computer simulations are presented in order to demonstrate that the proposed algorithm works.