Abstract: A dynamic state-space model (DSM) was presented for the problem of spectrum sensing under time-variant flat fading channels (TVFF). The primary user (PU) state and the fading channel were characterized as hidden states. On this basis, a novel method was proposed to estimate the PU state and the fading channel gain jointly and in real time. The simulation results show that the approach proposed could improve the performance of spectrum sensing apparently in time-variant flat fading channels.