Skip to Main Content
A Network of Kalman Filters (NKF)-based detector has recently been proposed for multiple access interference (MAI) and intersymbol interference (ISI) compensation in DS-CDMA systems. This paper presents a modified detector which deals with an "ε-contaminated" impulsive noise model. The model of the probability density function (pdf) of the noise as a sum of two weighted Gaussians is exploited in order to robustify the NKF detector. It is shown that the derived new version of the NKF can be viewed as a result of two NKF working on two different variances of background noise and coupled via the contamination parameter (ε). The resulting NKF detector is found to outperform the NKF detector optimised under the assumption that the measurement noise is Gaussian. A MAP symbol-by-symbol detector in impulsive noise is also derived for comparison.