Sparse coding based spectrum sensing in presence of multiple frequency hopping primary users

A novel sparse coding based wideband spectrum sensing technique for multiple frequency hopping spread spectrum primary users has been proposed. The proposed method relies on the sparsity of the received target signal on a learned exemplar dictionary. The learned exemplar dictionary is derived from training data comprising of undistorted primary user transmissions through all the channels in the band of interest using K-SVD algorithm. The high sampling requirement for the wideband sensing is addressed by using a set of lower rate samplers, each sampling the received baseband signal at different time offsets. The proposed method maintains constant false alarm rate under various SNR conditions and when the number of active primary users in the band is varying. On comparing with an existing FFT based wideband sensor, the detection performances of the proposed method for varying SNR and multiple active primary users are noted to be promising.