Blind adaptive stochastic neural network for multiuser detection

In this paper some blind adaptive methods are introduced for multiuser detection (MUD). The detector architecture contains a channel identifier followed by a stochastic Hopfield (1985) net. Blind channel identification is proposed to be carried out by either the Kohonen (see Self-Organizing Maps, Springer, 2000) algorithm or by a novel adaptive decorrelation technique. Based on the estimated channel parameters the stochastic Hopfield net implements a near optimal decision. Besides describing the related algorithms, the paper contains extensive simulations to evaluate the performance of the proposed detector structures.