Broadband Spectrum Sensing Algorithms Using SVM

Compressed sensing (CS) has been extensively used for spectrum sensing in cognitive radio (CR) system. However, in these CS-based spectrum sensing schemes, the reconstruction algorithms of compressed sensing cause high computational cost and bad detection performance in the low SNR. This paper proposes two compressed wideband spectrum sensing algorithms, which utilize Support Vector Machines (SVM) to establish spectrum detection classifiers instead of signal reconstruction and decision-making. The single-level multi-classifier sensing algorithm and the multi-level binary classifier sensing algorithm are designed respectively to meet the different requirements of real-time applications. The multi-level binary classifier sensing algorithm is suitable for scenarios with low real-time requirements which contain few spectrum channels. The single-level multi-classifier sensing algorithm with reduced computational complexity, is optimized for real-time detection systems. Simulation results show that the proposed algorithms outperform the classical energy detection algorithm based on CS. The proposed algorithms present a better robustness performance to noise and enhance the performance in the case of low SNR.