Blind least-squares approaches for joint data/channel estimation

This article addresses the problem of recovering blindly a source which has been sent through a multipath environment in a wireless multichannel context. A possible approach, primarily based on a joint data/channel estimation strategy, is outlined. The single-input/multiple-output (SIMO) deconvolution problem is first considered in a purely deterministic context, based on the minimization of a bilinear least-squares cost function, where the parameters to be adjusted are the channel coefficients and the transmitted signal vector, regardless of the finite alphabet property. A similar-output matching philosophy is used to construct a blind adaptive multichannel equalization scheme, with decision-feedback. The simulations show the robustness of the algorithm with respect to problems like channel order estimation errors and lack of channel diversity