Spectrum Sensing for Cognitive Radios in Time-Variant Flat-Fading Channels: A Joint Estimation Approach

Most of the existing spectrum sensing schemes utilize only the statistical property of fading channels, which unfortunately fails to cope with the time-varying fading channel that has disastrous effects on sensing performance. As a consequence, such sensing schemes may not be applicable to distributed cognitive radio networks. In this paper, we develop a promising spectrum sensing algorithm for time-variant flat-fading (TVFF) channels. We first formulate a dynamic state-space model (DSM) to characterize the evolution behaviors of two hidden states, i.e., the primary user (PU) state and the fading gain, by utilizing a two-state Markov process and another finite-state Markov chain, respectively. The summed energy, which serves as the observation of DSM, is employed for the ease of implementation. Relying on a Bayesian statistical inference framework, the sequential importance sampling based particle filtering is then exploited to numerically and recursively estimate the involved posterior probability, and thus, the PU state and the fading gain are jointly estimated in time. The estimations of two states are soft-outputs, which are successively refined with a designed iterative approach. Simulation results demonstrate that the new scheme can significantly improve the sensing performance in TVFF channels, which, in turn, provides particular promise to realistic applications.