Pathwise polynomial expansion interference cancellation for DS-CDMA

In DS-CDMA systems with time-varying multipath propagation, one of the main obstacles to linear multiuser detection is the large number of parameters which have to be estimated from scarce training data. Pathwise interference cancellation is an approach that allows us to separate the parameters into fast-varying and slowly varying parameters, thereby allowing the scarce training data to be used in the estimation of the fast-varying parameters with a short time constant while the slowly varying parameters can be estimated over a much larger time interval. We investigate the application of polynomial expansion (PE) to pathwise processing and propose the use of a weighting factor per signal component. We show that these weighting coefficients not only achieve significant improvements in the presence of power imbalances between users and paths w.r.t. scalar weighting, but also achieve further improvement due to the better estimation of the fastly varying parameters.