Adaptive linear detection for DS-CDMA communications in impulsive channels

Recently, there has been much work done on deriving adaptive linear detectors for DS-CDMA systems corrupted by additive Gaussian noise. However, such communication systems are often interfered with by noises other than the classical white Gaussian noise, and in here we consider DS-CDMA systems corrupted by natural impulsive noise sources, such as those found in low-frequency atmospheric channels, and for channels corrupted by man-made impulsive sources such as those occurring in urban or military radio networks. In this paper, we develop an adaptive linear detector, for such impulsive noise channels, which directly minimizes the average probability of bit-error, using an infinitesimal perturbation analysis (IPA) based stochastic gradient algorithm. The implementation allows a very simple recursive structure, and the conditions for convergence of this algorithm are presented. The performance of the detector is uniformly better than that of the linear correlator even under the extreme cases when either the multiple access interference or the impulsive noise is dominant.