This paper studies a scheme for blind source deconvolution where the source signals are mixed by a dynamical system. Under the assumption that the mixing system belongs to a given class, a learning law is given in order to recover the original signals.
