Channel Estimation Methods For Time- Varying Fading OFDM Channel Based On Channel Prediction And Decision-Feedback Equalization

Reliable channel estimation is the foundation for reliable transmission. The traditional channel estimation methods are either training sequence based methods or blind methods. The training sequence based methods appear to be robust methods but present some disadvantages. For example, bandwidth efficiency decreases as a non-negligible part of the data burst may be occupied for pilots. Blind methods allow the estimation of the channel based only on the received signal without any training symbols. The blind methods, however, require some special channel conditions and leave indeterminacy in channel estimation.We advance a novel method for the Single-Input-Single-Output (SISO) time-varying channel estimated of orthogonal frequency-division multiplexing (OFDM) systems. Firstly, the autoregressive (AR) models with proper order are used to approximate the time variation of the coefficients of wireless time-varying channel. And then, a Kalman filter based on the estimated AR model and past observations is used to predict (one-step ahead) the future channel. With the help of the predicted channel, we choose out the good DFE (Decision-Feedback equalization) points in the transfer function and then decide the most possible sent symbols, and they are used to reestimate the channel estimation and then to help to predict the future channel. Analysis and simulation results indicate that our method has a significantly good performance. The proposed method has as a closely similar effect as the piloted aided methods but needs much fewer pilots so that it improves bandwidth efficiency greatly. The estimated coefficients can also be used...