Spectrum Sensing for Cognitive Radios with Unknown Noise Variance and Time-variant Fading Channels

The unknown noise variance and time-variant fading channels make the spectrum sensing design a challenging task for cognitive radios (CRs). Most existing sensing methods suffer from the information uncertainty and can hardly acquire promising performances in the adverse situations. To address this challenge, in this paper, we firstly formulate a dynamic state-space model (DSM) for spectrum sensing, in which the unknown noise variance and time-variant flat fading channels are all taken into considerations. The dynamic behaviors of both primary user (PU) states and fading channels are characterized by two discrete state Markov chains. Based on this model, a novel spectrum sensing scheme is designed to recursively estimate the occupancy state of primary users, by estimating the timevariant fading channel gain and noise parameters jointly. The joint estimation is primarily premised on a maximum a posteriori probability (MAP) criterion and the marginal particle filtering (MPF) schemes. Simulation results are provided to demonstrate the advantages of our proposed method, which can significantly improve the sensing performance over time-variant flat fading channels, even with unknown noise variance.