Bayesian Learning for BPSO-Based Pilot Pattern Design Over Sparse OFDM Channels

In this paper, to investigate sparse channel estimation in OFDM communication systems, we propose a novel binary particle swarm optimization (BPSO) based pilot pattern design scheme and develop an efficient sparse Bayesian learning (SBL) scheme for sparse channel recovery. First, through modifying the mutual incoherence property (MIP) criterion, we outline a new penalty function for optimizing pilot pattern design, which comprehensively takes into account the overall coherence of the measurement matrix. Second, we modify the conventional BPSO algorithm by proposing a new adaptive inertia weight scheme, in which the inertia weight varies with the current state of particle swarm and the number of iterations. Furthermore, we map the pilot pattern design into the framework of the modified BPSO algorithm. Third, we develop a partitioned matrix iterative mechanism to compute matrix inversion in each iteration involved in SBL techniques. Finally, our numerical and simulation results illustrate the efficacy of the BPSO based pilot pattern design scheme and our proposed SBL algorithms in terms of mean-square estimation (MSE) error performance.