Improved semi-blind spectrum sensing for cognitive radio with locally optimum detection

In cognitive radio, there might be some information about primary users’ signals available at secondary users’ receivers since communications systems usually employ training signals for channel estimation and synchronization purposes. This training information can be exploited along with data symbols to perform semi-blind detection of primary users’ signals. In the literature, it is considered that the locally optimal semi-blind detection metric is the linear combination of the energy detector (ED) and the matched filter, i.e. the hybrid detector. Locally optimum detection (LOD), known to be optimum in the low signal-to-noise ratio, is proposed here in the design of a weighted semi-blind locally optimum detector (WSBLOD) by focusing on linear modulation in presence of an unknown phase shift and additive white Gaussian noise. By using LOD, it is shown that for binary phase shift keying-modulated signals, the semi-blind detector test statistic consists not only in combining linearly the matched filter and the ED but also the pseudo-energy of the received signal. Then, the designed semi-blind detector is improved by optimising the weights of the matched filter, energy and pseudo-energy in the test statistic, which maximises the probability of detection. Simulation results show that the proposed WSBLOD outperforms the hybrid detector.