Bayesian Matching Pursuit-Based Channel Estimation for Millimeter Wave Communication

Hybrid precoding is considered as a solution to reduce the high power consumption caused by devices operating at radio frequency (RF) in millimeter wave (mmWave) communication. For hybrid precoding, the channel state information (CSI) is critical but hard to obtain because of the analog precoding at RF and the large number of antennas. mmWave channel has been proved to be sparse by real-world experiments. Compressive sensing (CS) methods can be applied to the channel estimation to decrease complexity. However, there is a distinct performance gap between the estimation of the existing CS methods with or without given sparsity pattern (SP). In this letter, a new method based on Bayesian matching pursuit(BMP) idea is proposed to improve sparse channel estimation performance. We make appropriate assumptions according to the characteristics of mmWave channel. We select a set of candidate SPs with high posterior probabilities to estimate CSI. Numerical simulation shows that our proposed method has significantly improved channel estimation performance with acceptable complexity compared to existing methods including orthogonal matching pursuit, sparse Bayesian learning and Bayesian compressive sensing.