Semi-blind maximum likelihood separation of linear convolutive mixtures

We address the problem of separating a linear convolutive mixture of second order white sources, given some side information about the transmitted messages. The proposed technique exploits the special structure of the observed data matrix, after channel whitening: it is the product of orthogonal and generalized Toeplitz matrices in additive Gaussian noise. We implement the joint maximum likelihood (ML) estimator of both the orthogonal mixing matrix and the user signals, subject to the known algebraic and temporal constraints. Preliminary computer simulations assess the promising performance of the proposed method