Schnorr-Euchner Sphere Decoder with Statistical Pruning for MIMO Systems

A near-maximum-likelihood (ML) detection algorithm for spatially multiplexed multiple-input multiple-output (MIMO) systems has been considered. The sphere decoder (SD) is one of the promising techniques to solve the ML problem. However the SD has a loose necessary condition for pruning branches, and it becomes impractical in large dimensional systems. We propose a Schnorr-Euchner SD with statistical pruning (SP-SESD) in order to further reduce complexity with small performance degradation. Squared statistical constraint radius (SCR) and expected partial path metric from unvisited levels are defined from statistics of noises, and two pruning conditions are jointly applied to search tree for detection efficiency. A flexible trade-off between bit err-or rate (BER) and complexity can be supported by selecting two pruning probabilities in the proposed scheme, and hence one can design various MIMO detectors according to system demands. Simulation results show the proposed SP-SESD requires lower computational complexity than any statistical pruning approaches, although performance degradation is negligible. The proposed algorithm is effective for MIMO systems with any number of antennas.