Big-Data-Based Intelligent Spectrum Sensing for Heterogeneous Spectrum Communications in 5G

Abstract

Although spectrum sensing is commonly used in modern wireless communications to determine spectrum resources, the rapid development of wireless communications has generated massive heterogeneous spectrum data, which has dramatically increased the complexity of spectrum sensing. Machine-learning-assisted spectrum sensing, as an emerging and promising technique, provides an effective way to find available spectrum resources through the analysis of big spectrum data. In this article, a bigdata- based intelligent spectrum sensing method is proposed to improve heterogeneous spectrum sensing. Specifically, a cooperative spectrum sensing network is designed and established to realize wide-area broadband spectrum sensing and obtain big spectrum data. The effectiveness of such a network has been verified through detection probability simulation. To improve the reliability of spectrum sensing data, the correlations of the big spectrum data in time domain, frequency domain and space domain have been investigated, and the spectrum similarity has been obtained. Then a novel dual-end machine learning model is proposed to improve the precision and real-time prediction of heterogeneous spectrum states. Furthermore, a big spectrum data clustering mechanism is adopted to facilitate data matching and heterogeneous spectrum prediction. Finally, the comprehensive spectrum state is obtained through heterogeneous spectrum data fusion.