Joint Channel Estimation/Detection in MIMO Systems Using Belief Propagation

We study the problem of joint channel estimation and detection in MIMO systems using belief propagation (BP). We propose a BP based algorithm which jointly estimates the MIMO channel matrix and detects the data symbols in a frame (consisting of pilot and data symbols) through message passing in two passes; one pass from the start to the end of a frame and another from the end to the start. A message comprises a small number of parameters that characterize the computed densities of the marginal probabilities. We study the performance of the proposed algorithm in V-BLAST and STBC MIMO systems for varying pilot density and pilot/data power allocation in a frame. Our simulation results show that a close to perfect CSIR performance is achieved by the proposed algorithm using a non-uniform pilot/data power allocation scheme based on the knowledge of the average channel SNR. We show that the proposed BP based joint channel estimation/detection scheme performs better than MMSE channel estimation followed by maximum-likelihood (ML) detection in slow fading.