On error variance for autoregressive process-based spectrum occupancy prediction with energy detector for cognitive networks

Spectrum sensing is considered to be the most important task for both cognitive radio (CR) networks and next generation wireless networks (NGWNs). Spectrum sensing operation allows wireless nodes to identify vacant bands so that radio frequency (RF) spectrum could be used more efficiently. It is clear that a powerful spectrum sensing method equipped with an effective prediction strategy will carry utilization of the RF spectrum to the utmost level. Analysis of the performances of prediction strategies plays a critical role in quantifying the efficiency of the overall spectrum sensing operation. Therefore, in this study, autoregressive process-based prediction strategy for energy detection is investigated. It is shown that prediction error variance relies heavily on both the intercept of the autoregressive model and the threshold value selected. Theoretical findings are validated and verified by measurement data which are obtained by capturing the complete Global System for Mobile (GSM) downlink band at Nyquist rate of in-phase/quadrature (I/Q) level. Results and relevant discussions are provided along with future research directions as well.