科创中国

Sensing Matrix Design for MMV Compressive Sensing

• 摘要：

  Compressive sensing (CS) has been widely used in vehicular technology including compressive spectrum sensing, sparse channel estimation, and vehicular communications. The complete procedure of CS consists of sparse representation, sparse measurement, and sparse recovery. In the CS, measurement matrix Φ is utilized to sample a sparse signal, while sensing matrix Ψ is exploited for sparse recovery. It is important to properly design the sensing matrix that significantly affects the signal recovery performance. This paper addresses the issue of the sensing matrix design for the CS with multiple measurement vectors (MMV). We first study sufficient conditions for the MMV-CS and then develop three sensing matrix design approaches in different situations. The proposed methods have their roots in the principle of minimum variance distortionless response (MVDR) beamformer. This allows us to optimally determine the sensing matrix in the framework of the MVDR, which amounts to minimizing local cumulative cross-coherence between Φ and Ψ while keeping the columns of Ψ highly correlated to their counterparts of Φ. Simulation results are presented to verify the effectiveness of the proposed methods.

• 作者：

  Liang Zhang;Lei Huang;Bo Li;敬伟 殷;为民 包

• 刊名：

  IEEE Transactions on Vehicular Technology

• 在线出版时间：

  2019-9