Prediction of State Transitions in Rayleigh Fading Channels

This paper presents a channel-sampling scheme that allows robust estimation of channel transitions into and out of the fade state. The sampling scheme is based upon a second-order autoregressive (AR-2) model for the Rayleigh fading channel and a corresponding state-space representation in terms of fade and nonfade states. The threshold parameter that segments the fading signal into states is derived as a function of signal-to-noise ratio and a bit-error-rate performance metric. The sampling rates correspond to the values at which the mutual information of the process occupying a particular state, conditioned on two past observations of the channel, is maximized. The performance of the proposed sampling scheme in a model-based prediction algorithm is presented. The state of the received envelope is predicted with 98% accuracy. When applied to the estimation of channel gain, the sampling scheme in conjunction with a Kalman-filter-based AR-2 predictor yields one-step forecasts that accurately track the fading signal. Equalization of multipath channels using the estimated channel impulse response shows improved error performance over traditional recursive least squares equalizers