Energy detection techniques for signals subject to arbitrary channelizations

This paper considers the problem, in the context of cognitive radio systems, of identifying the spectral occupancy of a signal, where its bandwidth is only a subset of the total bandwidth of interest. It is argued that traditional detection approaches need to be extended to cater for arbitrary radio environments that are not channelized in a way that is known in advance. An extended energy detection model is introduced, where upon deciding that the signal is present, it also makes a decision on its spectral boundaries. A simple peak detection algorithm is described which allows these boundaries to be identified, and whose performance under various scenarios is observed empirically.