Next generation of CDMA wireless communication standards must support higher, more variable data rates. These requirements make the receiver more sensitive to multiple access interference(MAI), which limits the number of simultaneous users and their data rates. Multi-user detection (MUD) is a key technique for combating MAI. The high complexity of present MUD schemes has prevented its widespread implementation. This dissertation presents a set of candidate MUD schemes that offer a promising range of performance/complexity trades, for implementation in cellular base stations. We have considered partitioned reduced complexity techniques, where in multiuser interference suppression is shared between more than one processing block. Our scheme considers three such blocks, consisting of front-end processor, a group interference canceller (GIC), and a final reduced complexity trellis detector (RCTD).
For GIC, user signals are divided into groups, and partial interference cancellation is performed in a successive group-wise fashion. Performance of the technique shows SNIR gains of several, and concomitant improvements in error probability, with lower computational complexity than that of parallel or serial interference cancelling techniques. In addition, we determine optimal group sizes for our technique, where optimality is in terms of average error probability overall users.
We also present a new reduced complexity trellis detector. This detector, in near-far scenarios, utilizes hard decisions on the bits of the user signals with larger received energy for pruning away some of the possible states of the maximum likelihood sequence detector(MLSD) trellis, there by collapsing the trellis and reducing the computational complexity. We show that the proposed detector can achieve 50% reduction in computational complexity compared to MLSD with no loss in performance provided a moderate near-far ratio exists.
Results are presented for the combined multistage detector that depict performance and complexity improvement over the conventional single user detector and other MUD techniques, in an AWGN and Rayleigh fading multi-user asynchronous channel.