Adaptive nonlinear equalization for CDMA communication systems

We present an adaptive interference canceler in multiuser and multipath scenarios. The detector has two key elements: an adaptive algorithm for separating the signal subspace from the noise subspace and a decision feedback equalizer (DFE) based on the generalized cerebellar model articulation controller (GCMAC) neural network. The combination of the separator and the GCMAC-based DFE provides a close approximation to a Bayesian receiver. Simulations results show that the performance of the proposed receiver is comparable to that of the minimum mean-square error (MMSE) receiver, which is known to approach the single-user bound. An important advantage of the proposed receiver is the inherent capability to recompute its coefficients under decision-directed operation. In particular, the receiver does not require retraining when the adjustable weights deviate from their optimum setting