Spectrum Sensing in Cognitive Radio: Multi-detection Techniques based Model

Abstract

Cognitive radio (CR) paradigm is a new radio technology proposed to solve spectrum scarcity and underutilization. Central to CR is spectrum sensing (SS), which is responsible for detecting unoccupied frequencies. Since Detection techniques differ in their performance, selecting the optimal detection method to locally perform SS has received significant attention. This research work aims to enhance the reliability of local detection decisions, under low SNR, by developing a spectrum sensing that can take advantage of multiple detection techniques. This model can either select the optimal technique or make these techniques cooperate with one another to achieve better sensing performance. The model performance is measured with respect to detection and false alarm probability as well as sensing time. To develop this model, the performance of three detection techniques is evaluated and compared. Furthermore, the voting and the maximum a posteriori probability (MAP) fusion models were developed and employed to combine spectrum sensing results obtained from the three techniques. It is concluded that the cyclostationary feature detection technique is a superior detector in low SNR situations. MAP fusion model is found to be more reliable than the voting model.