Target sensing for cognitive radios

Since spectrum sensing provides the foundation of a cognitive radio system, it has to be highly reliable. Most often, the bandwidth of the underlying sensor is much smaller than the bandwidth of interest. Conventional swept sensing methods can easily miss primary user systems with either relatively low duty cycles or fast hopping. A traditional way to improve the detection probability in such situations is to sense and average over an appropriate time interval, but obviously at the expense of precious system resources. We suggest to use predictive modeling of the radio environment and sense only those spectral areas, which are expected to be occupied by a primary system. We name this novel algorithm `target sensing´ as the sensing approach is adapted to the target of interest. In this contribution, we discuss different challenges and issues involved with target sensing. Moreover, we compare its performance with conventional sensing methods. We can state that target sensing performs significantly better, especially when small sensing bandwidth platforms are used.