Joint Detection and Estimation on MIMO-ISI Channels Based on Gaussian Message Passing

For a bit-interleaved coded MIMO communication system over a frequency-selective channel we present a factorgraph-based joint channel estimation and data detection. In the graph, Gaussian message passing with different message update rules is investigated. We compare the application of belief propagation on the entire graph to a joint application of belief propagation in the detection region with either expectation maximization or mean field methods in the estimation region. In simulations, the performance in terms of frame error rate and estimation error is evaluated. While having similar computational complexity, we show that expectation maximization and mean field methods perform better for a low amount of available pilot information. For sufficient pilot information, all three methods provide similar results.