Background Spectrum Classification for Cognitive Radio

Recent changes in policy regarding the opportunistic use of licensed radio spectrum have paved the way for new innovative technologies like cognitive radio (CR). This technology puts tight demands on systems built to sense spectrum occupancy. Any strategy employed for opportunistic spectrum usage has to consider the tradeoffs between time spent searching for empty channels and time spent using those empty channels. In most cases the spectrum sensing that is employed by a CR system starts with no prior information about the occupancy of the channels it intends to use. A classifier can be run before the CRs attempt transmission to provide the CRs´ spectrum sensing sub-systems with a set of occupancy probability categories for some of the channels. By providing priors it may be possible for the CR to reach a transmission strategy in a shorter time frame. We propose a novel method of addressing this lack of prior knowledge by employing an efficient strategy that classifies some of the channels the CR intends to use within a fixed time limit. Our classification algorithm is based on multiple sequential probability ratio tests (multi-SPRT) and a heuristic allocation strategy for measurements that considers the completion time of each multi-SPRT. We will show that this strategy will achieve a bounded error by prioritizing channels that give consistent measurement results. We also compare the performance of the proposed system to simpler systems that do not require as many computations.