A review of spectrum sensing in modern cognitive radio networks

Abstract Cognitive radio network (CRN) is a pioneering technology that was developed to improve efficiency in spectrum utilization. It provides the secondary users with the privilege to transmit on the licensed parts of the spectrum if the licensed user is not utilizing it. The cognitive radio must, however, relinquish the spectrum when the primary user decides to reoccupy it. By exploiting the unused portion of the spectrum, a cognitive radio helps in making the use of the radio spectrum more efficient. Furthermore, the most important capability that a cognitive radio (CR) must possess is spectrum sensing. A CR must be able to correctly determine the status of the target spectrum with the help of spectrum sensing. This is a very challenging task and several methods have been investigated over the years. In this work, the state of the art of different spectrum sensing techniques for a variety of CRNs is presented. Both conventional and modern spectrum sensing techniques for different types of primary user signals are discussed in this work for Narrowband and Wideband signals. Legacy techniques such as energy detection are most commonly used due to their simplicity in implementation. However, this comes at the cost of poor performance at low SNR (signal-to-noise ratio) values. This issue is countered by methods that use statistical information of the primary signal to make a more informed decision on spectrum occupancy. Several techniques that make use of the power of machine learning algorithms are also discussed which show clear improvement in performance. The primary challenge in such techniques is selection of the best features. The most commonly used features are also discussed. Furthermore, spectrum sensing techniques that consider the 5G signal as the primary user signal of the network are discussed. It is observed that there is a significant need for research in additional spectrum sensing techniques for 5G cognitive radio networks.