Graph-Based Soft Channel and Data Estimation for MIMO Systems with Asymmetric LDPC Codes

Abstract

In this paper, we propose an iterative soft channel estimation and data detection algorithm based on a factor graph. Channel coefficients as well as data symbols are treated as variable nodes and are all estimated in a low-complexity element-wise manner. Applying asymmetric LDPC codes, this algorithm is able to deliver ambiguity-free outputs for MIMO systems with or without training symbols. Training symbols are inherently utilized as a type of a priori information. This algorithm thoroughly relaxes the troublesome constraints on training design in the sense that an arbitrary (even zero) number of training symbols can be placed at arbitrary positions within a data burst.