Parallel Interference Cancellation - a Multiuser Detection Framework

Parallel interference cancellation (PIC) principle is conceptually simple and yet eeective method of parameter estimation and data detection in the presence of multiple-access and multipath interference. It uniies some of the known methods of data detection, complex channel co-eecient estimation, as well as delay tracking in DS-CDMA receivers. The PIC method will be derived and some numerical results for data detection, channel coeecient estimation and delay tracking will be presented in the paper. A great deal of the recent research on multiuser CDMA receivers addresses the problems of multipath fading. An optimal receiver requires joint estimation of complex channel coeecients, propagation delays and data. Especially when tracking time-variant channel parameters, the multidimen-sional optimization problem can be approximated by KL separate estimation problems. The other user parameters are assumed to be known when estimating the kth user's lth path delay and complex channel coeecient. This leads to parallel interference cancellation (PIC) in channel parameter estimation: an estimate of the multiple access interference (MAI) is subtracted from the matched lter outputs. Parallel interference cancellation is conceptually simple and yet eeective method of parameter estimation in the presence of MAI. It uniies some of the known methods of data detection, complex channel coeecient estimation, as well as delay tracking. In data detection, parallel interference cancellation is used in the multistage algorithm 1] at each cancellation stage. The PIC based delay tracker is quite similar to the expectation maximization (EM) delay tracker presented in 2]. The EM algorithm has been used in 3] for estimating the complex channel coeecients in a multi-stage receiver. The recursive algorithm developed in 3] is equivalent to the PIC based channel estimator presented in this paper. The recursive EM algorithms presented in 2, 3] have been used for estimating xed unknown channel parameters. The EM and PIC algorithms have fundamental diierencies when applied to time-variant parameter estimation problems. As a matter of fact, the basic principles of EM algorithms 4] will be violated in their recursive implementations when applied to time-variant channel parameter estimation 5, 6]. For that reason PIC methods will be introduced for estimating time-variant channel parameters. The performance of the PIC based algorithms depends heavily on the quality of MAI estimates, which are formed by using channel coeecients, data, and delay estimates of all users. Degradations in any of these parameters impairs the performance of PIC based algorithms. Especially inaccurate complex channel coeecient estimates have a …