Adaptive multiuser demodulation for wireless multiple access communications

Adaptive multiuser demodulation of digital signals in multiple-access cellular communication systems is considered. The goal of this research is to develop and evaluate receiver algorithms which will provide reliable estimates of the source symbols transmitted by all users of interest. The emphasis of our study is on adaptive multiuser demodulation, i.e., the receivers are able to adapt to time-varying channel conditions and interference. In addition, we consider scenarios where training data may not be readily available and/or near-far effects are likely. The wireless multiple-access communication systems considered include Time-Division Multiple Access systems such as GSM, and Direct-Sequence Code-Division Multiple Access (DS-CDMA) systems. We evaluate several suboptimal receivers and their capabilities as well as limitations.;We first consider linear receivers in combination with spatial diversity. These include the Minimum Mean-Squared Error receiver, the adaptive Least-Squares (LS) receiver, and reduced-rank receivers with and without training symbols. Numerical results show that reliable reception is possible even when near-far effects are severe. The proposed blind receiver performs well when the number of antennas is sufficiently large and when users differ substantially in received power.;Next, we consider another suboptimal group of receivers, the Decision Feedback Demodulator (DFD). The DFD promises superior performance to the linear receivers at the cost of a moderate increase in complexity. We study both the Successive DFD (S-DFD) and the Parallel DFD (P-DFD). Salient features of the proposed DFD include the ability to cancel interference without remodulation, simple adaptive implementation, and reduced computational complexity as compared with the optimal receiver. Adaptive estimation algorithms based on the Least-Squares criteria are presented. Simulations show that a substantial performance improvement in BER relative to the linear LS receiver can be obtained.;Finally, we study the average probability of error for both the S- and P-DFDs in a DS-CDMA system where the average is over randomly assigned signature sequences. We present a large system analysis which allows an efficient characterization of receiver performance as a function of various system parameters. The results show that at low load and/or high Signal-to-Noise Ratios, the probability of error approaches single-user performance.