Perceptron-based CDMA receiver in non-Gaussian noise

Linear adaptive interference suppression is an effective approach in mitigating Multiple Access Interference (MAI) in CDMA communications channels under Gaussian noise. However, applying Gaussian-based linear receivers to realistic communications channels affected by impulsive noise can result in performance degradation and is therefore a suboptimal solution. The linear adaptive receiver is limited by its structural formulation as a simple linear combiner with a hyperplanar decision boundary that is unable to form arbitrarily non-linear boundaries inherent in CDMA communications, particularly under hostile channel conditions. In this paper, we consider the use of a multilayer perceptron (MLP) for robust single-user detection in multiuser non-Gaussian channels; the receiver is trained via the back-propagation (BP) algorithm using a known training sequence. Theoretical arguments supported by simulation suggest that the perceptron-based receiver, which combines the basic attributes of nonlinearity, adaptivity and non-parametric approach, shows promise as a near-far resistant receiver in non-Gaussian environments.