COOPERATIVE SPECTRUM SENSING USING MACHINE LEARNING-BASED MODELS

Abstract

The unstoppable evolution that has affected mobile telecommunication systems in the last three decades has caused the occupation of the licensed frequencies, but at the same time these frequencies are not being used efficiently. Cognitive Radio is the key technology introduced to overcome the main problems of the spectrum utilization, since it offers the opportunity for other unlicensed users to utilize the licensed band while it is not being used by primary user. Even though it increases the efficiency of spectrum utilization, spectrum sensing in cognitive radios still faces problems for higher-performance and more energy-efficient systems. In this work, are taken in consideration two machine learning algorithms as decision-making tools in the fusion centre of cooperative spectrum sensing network based on energy detection technique. The effectivity of these algorithms is evaluated using Receiver Operating Characteristics (ROC) curve and Area Under The Curve (AUC) values, considering seperately additive white Gaussian noise and Rayleigh fading channel. Moreover, the training period of each algorithm is analyzed to evaluate the execution cost for each of them.