Iterative Equalization Algorithm For Double Selective Fading Channels

Due to the existence of the Doppler shift and the multi-path propagation,the signal transmission over the doubly-selective fading channels may encounter the problems of fastvarying channel state information and inter-symbol interference.These problems may further undermine the detection accuracy at the receiver.The equalization technique can alleviate the inter-symbol interference problem at the receiver.However,existing equalization algorithms fail to address the problem of fast-varying channel state information.In addition,exsiting solutions fail to achieve an efficient trade-off between the computational complexity and the detection accuracy.In order to address these problems,we consider the equalization task over the doubly-selective fading channels,and our work is summarized as follows.In order to improve the detection accuracy and detection efficiency over the doubly-selective fading channel,we propose an iterative receiver design with Linear Minimum Mean Squared Error(LMMSE)based pre-processing and Gaussian Approximation(GA)based equalization.In order to improve the estimation accuracy of the fast-varying channel,we propose the decision feedback strategy and the channel interpolation strategy,so that the receiver can dynamically track the doubly-selective fading channel.In order to improve the detection efficiency at the equalizer,we propose the LMMSE based pre-processing,so that more accurate prior information is provided to the GA based equalizer.The simulation results show that the proposed channel estimation strategies can effectively improve the estimation accuracy.Furthermore,the LMMSE pre-processing could significantly accelerate the convergence of the iterative GA equalizer.In order to improve the accuracy of Gaussian approximation as well as the convergence of the iterative LMMSE-GA equalizer,we adopt the Expectation Propagation(EP)principle and propose the EP-based iterative LMMSE-GA(LMMSE-EPGA)equalizer.Specifically,the conventional GA equalizer directly perform moment matching on the extrinsic information from the decoder,which results in lower approximation accuracy.To address this problem,we employ the EP principle,and perform moment matching on the full information for higher approximation accuracy.Simulation results show that the EP principle could effectively improve the detection accuracy as well as the convergence of the iterative LMMSE-GA equalizer.