Effect of Spectrum Occupancy on the Performance of a Real Valued Neural Networl Based Energy Detector

In this paper, a newly proposed Real Valued Neural Network (RVNN) based Energy Detector (ED) is presented for Cognitive Radio (CR) application. With little available on the performance of EDs in varying spectrum occupancy conditions, we provide a study to understand how occupancy variation affects the performance of a newly proposed RVNN based ED and other ED schemes. Other factors such as varying Signal to Noise Ratio (SNR) and model order values were also examined in this study and result analysis conducted using the Precision-detection statistics. Implication of results obtained indicate that the RVNN based ED would perform optimum in high occupancy and SNR conditions for a model order choice of P = 20. We also observed that the RVNN based ED would provide better precision performance characteristics over the Periodogram, Welch and Multitaper based ED schemes compared herein. Hence, the RVNN based ED suffices as a favourable choice for CR application even under varying occupancy conditions.