Antenna Cross-correlation based Compressive Subspace Learning for Wideband Spectrum Sensing.

To address the performance degradation of wide-band spectrum sensing by sub-Nyquist sampling (SNS) in wireless fading channels, two compressive subspace learning (CSL) algorithms are proposed for signal subspace learning based on antenna cross-correlations for further improving the sensing performance. Both algorithms are developed based on different organizations of SNS samples, and both exploit space diversity and noise uncorrelations between antennas. We further establish the expressions for statistical covariance matrices (SCMs) obtained by SNS samples in the multi-antenna SNS cognitive radio system. Based on the derived SCM expressions, conditions to ensure SCMs without noise contamination are given. Simulations validate the derived conditions and show the improvement on sensing performance over related works.