Spectrum sensing in cognitive radio using quickest change detection framework

Abstract

Spectrum sensing is a term associated with detection of primary (licensed) users (PU) by secondary (unlicensed) users (SU) to pursue opportunistic transmission of their data. The problem of spectrum sensing is challenging because of the distributed nature of SU. Since the presence or absence of PU has to be detected as quickly as possible therefore we use an approach based on quickest detection. To improve sensing efficiency we assume the SU to be equipped with multiple antennas for spectrum sensing. We also assume that each antenna makes fixed number of observations which are used to compute energy metrics. The energy metrics computed at various antennas are then combined using weights determined according to Fisher linear discriminant criterion. In our work we proposed a theoretical framework for change detection of the two hypotheses namely presence (or) absence of PU. From our simulation results we observe that the average detection delay decreases with an increase in the number of observations. The proposed weighted gain combining (WGC) gives lower average detection delay than the equal gain combining (EGC).