Adaptive Decision-Directed Quantized-State Algorithms for Multi-user Detection of CDMA Signals

The paper presents novel techniques for the multi-user detection of the CDMA signals wherein the communication channel may be a fading channel and the transmission from various users may be synchronous or asynchronous. The proposed architecture is based on a decision-directed approach and does not require any training sequence. The architecture involves a bank of matched filters whose outputs are input to an adaptive weighted combiner whose weight vector is determined by an adaptive algorithm. The class of algorithms investigated for the adaptation of the weight vector includes the well known recursive least squares (RLS), the least-mean square (LMS) algorithms, and the class of algorithms invented by the first author of this paper, and termed quantized-state (QS) algorithms. Various simulation results are presented in terms of the performance of the proposed architecture in both the fading and non fading environments showing performance close to that obtained in the multiple access interference free environment even when severe multiple access interference environment exists.