Towards the Performance of ML and the Complexity of MMSE - A Hybrid Approach

In this paper, we present a near ML-achieving sphere search technique that reduces the number of search operations significantly over existing sphere decoding (SD) algorithms. While the SD algorithm relies only on causal symbols in evaluating path metric, proposed method accounts for the contribution of non-causal symbols with the aid of per-path minimum mean square error (MMSE) symbol estimation. The ML and MMSE combined cost metric results in the tight necessary condition for sphere decision and hence expedites the pruning of subtrees unlikely to be survived. From the simulations performed over multi-input multi-output (MIMO) wireless channels, it is shown that the computational complexity of the proposed approach is substantially smaller than the existing SD algorithms while providing negligible performance loss.