Reliable channel estimation based on Bayesian compressive sensing for TDS-OFDM systems

Time domain synchronous OFDM (TDS-OFDM) has higher spectrum efficiency than standard cyclic prefix OFDM (OFDM) by replacing CP with a known training sequence as the guard interval of OFDM data block, but severe mutual interferences will be caused in multipath channels. Recent studies have shown that the theory of compressive sensing (CS) can be efficiently applied to achieve reliable channel estimation to solve this problem, but the CS-based channel estimation suffers from obvious performance loss when the channel sparsity is unknown or under or the signal-to-noise ratio (SNR) is low. In this paper, we propose the Bayesian compressive sensing (BCS) based channel estimation algorithm to solve these problems, whereby some prior information of the channels can be exploited to improve the performance when channel sparsity is unknown. Besides, we also combine the statistical learning theory (SLT) and the basic thoughts of relevance vector machine (RVM) to further improve the noise-resistibility of channel estimation when SNR is low. Simulation results indicate that the proposed BCS-based channel estimation algorithm can effectively solve the major problems of the traditional CS-based schemes.