Low-Complexity Sparse Bayesian Learning for Joint Channel Estimation and Target Detection in ISAC

Abstract

In massive multiple-input multiple-output (MIMO) Integrated Sensing and Communications (ISAC) systems, the joint task of high-dimensional channel estimation and multi-target detection suffers from a severe computational bottleneck, which limits practical deployment and system scalability. To address this problem, this paper proposes a low-complexity covariance-free sparse Bayesian learning (CoFe-SBL) algorithm. First, we establish a unified sparse signal model for beamspace communication channel estimation and target detection by leveraging the inherent sparsity of the scattering environments and the targets in the millimeter-wave frequency band. Then, we introduce the diagonal estimation principle to reformulate the SBL framework and combine it with a conjugate gradient (CG) solver and preconditioner optimization, enabling efficient estimation for each task. The simulation results demonstrate that the proposed CoFe-SBL algorithm achieves channel estimation accuracy and target detection performance comparable to existing SBL methods, while reducing computational complexity by a factor of nearly 3.5.