Applying Classification Methods for Spectrum Sensing in Cognitive Radio Networks: An Empirical Study

Abstract

Spectrum sensing is the paramount aspect of cognitive radio network where a secondary user is able to utilize the idle channels of the licensed spectrum band in an opportunistic manner without interfering the primary (license) users. The channel (band) is considered to be idle (free) when primary signal is absent. The channel accessibility (free) and non-accessibility (occupied) can be modeled as a classification problem where classification techniques can determine the status of the channel. In this work supervised learning techniques is employed for classification on the real-time spectrum sensing data collected in test bed. The power and signal-to-noise ratio (SNR) levels measured at the independent CR device in our test bed are treated as the features. The classifiers construct its learning model and give a channel decision to be free or occupied for unlabelled test instances. The different classification technique’s performances are evaluated in terms of average training time, classification time, and F1 measure. Our empirical study clearly reveals that supervised learning gives a high classification accuracy by detecting low-amplitude signal in a noisy environment.