Analysis of Machine Learning Algorithms for Spectrum Decision in Cognitive Radios

Abstract

Technological advances in recent years have reduced the manufacturing costs of wireless devices, increasing the number of such devices and applications. Most of these applications are supported by ISM (Industrial, Scientific, and Medical) frequencies, which due to their wide use in several types of devices have suffered from harmful interference. To solve this problem, Cognitive Radios paradigm has been proposed to guarantee the quality of communication. Several frameworks were proposed for the development of a Cognitive Radios Networks (CRN), but none of them were effectively implemented in hardware. This paper presents an analysis of machine learning algorithms in architecture for the development of CRN in real hardware. Results demonstrated the feasibility of the architecture and the decision methods based on machine learning algorithms can find the best communication channel.