Compressive slow-varying wideband power spectrum sensing for cognitive radio

Wideband spectrum sensing is a critical component of a functioning cognitive radio system. Its major challenge is the too high sampling rate requirement. Compressive sensing (CS) promises to be able to deal with it. Nearly all the current CS-based compressive wideband spectrum sensing methods exploit only the frequency sparsity to perform. This paper sets up a new signal model which is sparse in both temporal and frequency domain. Motivated by the achievement of a fast and robust detection of the wideband spectrum change, total variation minimization is incorporated to exploit the temporal and frequency structure information to enhance the sparsity level. As a sparser vector is obtained, the spectrum sensing period would be shortened and sensing accuracy would be enhanced. Both theoretical analysis and numerical experiments demonstrate the performance improvement.