Linear estimate-based look-ahead path metric for efficient soft-input soft-output tree detection

In this paper, we propose a new path metric, which improves performance of soft-input soft-output tree detection for iterative detection and decoding (IDD) systems. While the conventional path metric accounts for the contribution of symbols on a visited path due to the causal nature of tree search, the new path metric reflect the contribution of unvisited paths using an unconstrained soft estimate of undecided symbols. This path metric, referred to as a linear estimate-based look-ahead (LE-LA) path metric is applied to a soft-input soft-output M-algorithm that finds a list of promising symbol candidates and computes a posteriori probability of each entry of the symbol vector using the candidate list found. Through the analysis of a probability of correct path loss (CPL) and computer simulations, we show performance gain of the LE-LA path metric over the conventional path metric.