Unsupervised Learning for Robust Signal Classification

Abstract

Cognitive Radio Networks (CRNs) have been proposed to increase the efficiency of channel utilization; presently the demand for wireless bandwidth is increased. Cognitive Radio Network can enable the sharing of channels among unlicensed and licensed users on a non-interference basis. An unlicensed user (i.e., secondary user) should monitor for the presence of a licensed user (i.e., primary user) to avoid interfering with a primary user. However to get more gain, an attacker also called self-ish secondary user may copy a primary user’s signal to cheat other secondary users. Therefore a primary user detection method is needed to detect the difference between a primary user’s signal and secondary user’s signal. In this paper, unsupervised learning methods such as K-means and SOM techniques are used to classify the signals and also measure the performance parameters such as throughput, end-to-end delay, energy consumption, packet delivery ratio and collision rate of the channel.