Interference-aware time-frequency based spectrum sensing for cognitive radio networks

Time-frequency (TF) based cognitive radio (CR) sensing schemes have to operate in different environments. In case of completely known distributions of observations under both primary user (PU) present and PU absent hypotheses, sensing can be based on a conventional likelihood ratio testing approach. However, if the observations contain unknown interference signals the conventional approach has to be modified. In this paper, we derive suitable filter bank (FB) based TF sensing schemes where the non-Gaussian co-channel interference (CCI) at the FB output is observed to follow a complex symmetric a-stable distribution. Based on the model, we consider a generalized likelihood ratio test (GLRT) for detecting a PU signal in interference plus noise and analyzing the robustness of the FB against the CCI. In order to construct the test, the parameters of Gaussian mixtures under both hypotheses are estimated using an expectation-maximization approach and employed in the GLRT. Simulations are carried out for different relative intensities of the interfering signals and the sensing performance is characterized by receiver operating characteristics for the different environments.