Factor-Graph-Based Detection Algorithms for Coded OFDM over Time-Varying Channels

In this paper we consider a coded orthogonal frequency division multiplexing (OFDM) system over time-varying channels. We propose soft-input soft-output detection algorithms based on a cyclic factor graph representation of the a posteriori distribution of the transmitted symbol sequence. By applying the sum-product algorithm (SPA) to this graph we compute the a posteriori probabilities of each transmitted symbol, where the message schedule for the SPA determines the performance of the detection algorithms. Therefore we present on the one hand a serial schedule which provides a good bit error rate (BER) performance at the initial iteration but has a high latency and on the other hand a parallel schedule with low latency but a poor initial performance. Moreover we introduce a novel hybrid message schedule which provides a tradeoff between the aforementioned message schedules. The proposed detection algorithms provide a remarkable complexity reduction compared to the optimal forward-backward algorithm (FBA). By computer simulations of an OFDM-LTE downlink system with turbo equalization we demonstrate the effectiveness of the proposed algorithms, for instance we show that after 3 iterations the BER performance of the proposed algorithms is similar to that of the FBA.