Cooperative Spectrum Sensing for PU Detection in Cognitive Radio Using SVM

In the world of cognitive radio, spectrum sensing is among the challenging and essential tasks. For dynamic allocation of spectrum, the primary task is the detection of the primary user’s (PU) presence over the spectrum. Primary users can be detected through a single node (CR user) but there are various limitations in this method like multipath fading, hidden terminal problem and shadowing. For overcoming these limitations, cooperative spectrum sensing was proposed. In this paper, a cooperative spectrum sensing (CSS) based on the support vector machine (SVM) is being proposed. Firstly, we have generated a data set through energy detection method containing energy vector of a signal as a feature. Random over sampler is further used in order to balance the classes. After balancing, we have used various machine learning algorithms and made comparison between them to classify if a primary user is present over the spectrum. The comparison clearly shows that the best accuracy is given by the proposed SVM algorithm for classifying the presence of primary user. Simulation results also show the better capability, robustness and superior efficiency of SVM method as compared to other algorithms over the detector which we have used in this paper.