Wireless Fading Channel Estimation And Detection Based On Message Passing Algorithms

In wireless communication,precise channel state information is the prerequisite for equalization and detection of data symbols in receiver.However,high-mobility of terminal will result in the Doppler frequency shift,yielding the time-varying behavior of wireless channels;the symbol period is far less than channel delay spread,leading to the frequency-selective characteristic.The above two phenomena will reduce the accuracy of channel estimation and data detection.Therefore,how to design joint channel estimation and symbol detection algorithms with high performance and low complexity becomes the focus of researchers.Message passing algorithms based on factor graph and variational inference are effective tool for iterative signal processing,which can not only realize the unification of algorithms on probability graph model,but also overcome the problem of poor performance or high complexity of single algorithm.This thesis designs joint channel estimation and symbol detection algorithms by using factor graph as analysis tool and message passing algorithms as theoretical base.Firstly,this thesis discussed the significance and background of the topic,analyzed the research status of the formalized iterative receiver and message passing algorithms,introduced the wireless fading channel,the Bayesian estimation and variational inference,and showed the derivations of three kinds of message update rules.Secondly,considering unknown time-varying flat-fading channel,we design a formalized iterative receiver based on message passing.We obtain the system factor graph by factorizing the global posterior probability density function and then derive the messages passing between each node.Expectation propagation(EP)is adopted to deal with the high complexity caused by Gaussian mixture nodes.After that,we proposed a combined belief propagation(BP)and EP receiving algorithm and a suitable message scheduling mechanism to improve the performance.The simulation results show that the proposed algorithm outperforms the sum product particle message(SPPM)receiver in terms of bit error rate(BER),normalized mean square error(NMSE)of channel estimation and convergence rate at low complexity.Finally,considering unknown time-varying frequency selective fading channel,we proposed an BP-EP-MF(mean field)-based iterative receiver to realize joint channel estimation,equalization and symbol detection.A suitable message scheduling mechanism is designed and MF rule is adopted to cope with the non-closed solution issue caused by BP.Simulation demonstrates that the proposed algorithm could achieve better BER,NMSE of channel estimation and convergence performance comparing with extended Kalman filter receiving algorithm even with the slight decreasing of complexity.