Blind deconvolution of multivariate signals: A deflation approach

It is well established that the blind deconvolution problem makes sense as soon as y (or, equivalently, w) is non Gaussian. A previously adaptive blind deconvolution problem is discussed in the case where the observation y is a p-variate signal given by a particular function where the impulse responses (h_k)_{k = 1,p} are unknown p × 1 transfer functions, and where the excitations (w_k)_{k = 1,p} to be reconstructed are statistically independent non-Gaussian centered i.i.d. sequences of variance 1