Per-survivor interference cancellation structures for low-complexity ML detection of DS/CDMA signals

In this paper we investigate reduced-complexity approximations of ML multiuser sequence detection of DS/CDMA signals where the likelihood computations are based on reduced-state trellises. The proposed structures rely on survivor information to cancel interference terms in the branch metric equations. The implementation complexity of this family of detectors can be easily traded for performance, with structures ranging from the optimum ML detector to a successive, symbol-by-symbol interference cancellation scheme. Simulation results show a significant complexity reduction with marginal performance degradation