Adaptive based machine learning approach for cooperative energy detection in cognitive radio networks

Abstract In cognitive radio networking, spectrum can be utilized by a secondary user while insuring no interference to the primary user of the spectrum. This helps enhancing the utilization of the spectrum while considering the rights of its primary users. Secondary users need to actively detect the existence/absence of the primary user to deploy a cognitive radio network. By cooperating, secondary users can enhance the detection capabilities, especially in environments with fading and noise, thereby increasing the reliability of spectrum sensing. The objective of this work is to employ machine learning with feature extraction and random forest classifier to enhance the individual secondary user energy detection accurateness in presence of a high level of noise power density. Clustering method is used to organize the secondary users for cooperative decision making on the existence of the primary user. The detection probability is analysed based on the ROC, where it reaches approximately 0.95 at a probability of false alarm of about 0.05, indicating a highly efficient detection capability.