Time-varying channel estimation based on dynamic compressive sensing for OFDM systems

Time domain synchronous OFDM (TDS-OFDM) is more spectrum-efficient than standard cyclic prefix OFDM (CP-OFDM), but contemporarily it cannot support higher-order constellations like 256QAM due to the residual mutual interferences between the pseudorandom noise (PN) guard interval and the OFDM data block. An efficient way is to incorporate the compressive sensing (CS) theory into TDS-OFDM which utilizes inter-block-interference (IBI)-free regions of very small size to reconstruct the parse channel of high dimension, whereby a static sparse channel plays a significant role. In this paper, to handle more practical dynamic channels which are time-varying over symbol by symbol, we exploit the dynamic compressive sensing (DCS) and Bayesian inference to improve the reconstruction accuracy of dynamic channel estimation. Specifically, we realize a causal low-complexity filter using message passing (MP) algorithms in a factor graph to reconstruct a dynamic sparse channel by utlizing the channel property that path delays of dynamic channels change slowly and path gains vary smoothly over time. Simulation results show that the proposed approach outperforms previous static CS schemes over dynamic channels in practice.