Effect of uncertainties in modeling interferences in spectrum sensing

Summary form only given. Spectrum sensing is one of the major functions of cognitive radio. Its goal is to detect and classify spectrum holes, i.e., regions of the spectrum space that can be used by secondary users. A realistic assumption with the design of a spectrum sensor is that the channel model is incompletely known, i.e., only partial knowledge is available about the observation statistics. Under these conditions, one may consider using a detector designed under the assumption of a known model, and examining the loss in performance caused by the modeling uncertainties. One may suppose for example that the additive noise has a known statistics, except for its power which ranges in a certain interval. In this case, it is shown in that model uncertainties impose fundamental limitations on sensing performance, and these limitations cannot be overcome by increasing the sensing duration. This approach leads to the concept of SNR wall, which is the SNR level below which a spectrum sensor fails to yield an acceptable detector performance, no matter how long it can observe the channel.