Improved-MUSIC Algorithm Based on Compressive Subspace Learning for Multiantenna Cognitive Radio

Recently, cognitive radio (CR) has become an effective method for wideband spectrum sensing (WBSS) based on sub-Nyquist sampling to enhance the sensing range as much as possible and alleviate the spectrum resource constraint. However, most of the previous work tends to ignore the spatial correlation between channels, making the sensing performance degraded. The few works that consider spatiality often suffer from excessive computational complexity and overhead. In this paper, we consider space-dependent MIMO channels and propose two compressive subspace learning (CSL) algorithms (mCSL-improved MUSIC and vCSL-improved MUSIC). Both algorithms exploit spatial diversity, with mCSL exploiting the antenna averaged temporal decomposition and vCSL exploiting the spatial-temporal joint decomposition. By the singular value property of the statistical covariance matrix (SCM), we present the analysis of the MUSIC algorithm under two structures. To further reduce the computational complexity, we demonstrate the feasibility of improving the cut-off conditions of MUSIC based on the orthogonality of signal and noise, and give the mCSL-improved MUSIC and vCSL-improved MUSIC algorithm procedure and calculate the complexity. The proposed algorithm can ensure the performance of wideband spectrum sensing in the presence of noise while also reducing the computational complexity. Simulation results demonstrate the correctness of this work and its performance improvement of wideband spectrum sensing over related works.