A new particle filtering detector is proposed for joint estimation of channel model coefficients, channel tracking, and signal detection over flat Rayleigh fading channels. The detector employs a hybrid importance function and a mixture Kalman filter which result in a highly efficient implementation. In addition, by considering the practical limitation of the system and physical interpretation of the adopted AR(2) channel model, we realize a fully blind particle filtering implementation. Simulation are provided to show the performance of the proposed detector.