A Machine Learning based Spectrum-Sensing Algorithm

During the session dedicated to the Multimedia Communications & Smart Networking, at CHINACOM 2015, Haozhou Xue and Feifei Gao presented their paper titled ‘A Machine Learning based Spectrum-Sensing Algorithm Using Sample Covariance Matrix’, which was awarded as one of the six Best Papers of the conference.

The rapid development of wireless communication technologies and the low spectrum usage in many frequency bands, drove many studies around spectrum resources, with a special focus on the problem of spectrum scarcity. Within this context, cognitive radio (CR) appears as a promising technology for exploiting the underutilized spectrum in an opportunistic manner, and allowing secondary users (SUs) to share the licensed spectrum of primary user (PU) when the latter is not subject to harmful interference.

The state-of-the-art Current spectrum sensing methods include the matched filtering detection, the cyclostationary detection and the energy detection, which is the most effective one so far, due to the minimum priori information required. However, energy detection is vulnerable to the noise uncertainty and is not ideal for detecting the correlated signals.

A new approach Haozhou Xue and Feifei Gao propose a new spectrum sensing approach based on machine learning and the covariance matrix of the received signals received from multiple antennas. The research shows that this method performs well when the signals received by SU are highly correlated. Furthermore, it can be used in various signal detection applications without any prior information about the signal, the channel, and the noise power. In short, the proposed model is applicable when there is no knowledge of signal and environments, due to the beauty of the intelligent learning.

To read the full text of the paper, please visit the European Union Digital Library (EUDL).