Research On Signal Detection And Channel Estimation In Multi-antenna Communication Systems

In recent years,due to the rapid development of the data services such asmultimedia and Intemet,the higher data transmission rate has been required for futurewireless communications in order to meet the people's growing demand.Currently,there are a variety of techniques that can be used as the key candidates for futurewireless communications,among which Mulitple-Input Mulitple-Output (MIMO)technology gives direction to the development of the future broadband high-speedwireless data communications technology since it can dramatically improve the channelcapacity and spectrum efficiency without increasing the power and bandwidth.We firstly survey the research status of the signal detection and channel estimationin MIMO system and point out a variety of important problems and questions needed tobe solved.Then the main research contents and contributions of this dissertation areoutlined.In the second part of the dissertation,at the basis of the modifiedQR-Decomposition (QRD) with column pivoting,an adaptive group detectionalgorithm based on channel quality is proposed by combining the sphere decoding (SD)algorithm and the layer-by-layer detection method for Vertical Bell Laboratories SpaceTime (V-BLAST) system over an i.i.d.Rayleigh flat-fading channels.Very flexibletradeoff can be achieved between performance and complexity via selecting the groupthresholds.More significantly,the proposed algorithm unifies the SD algorithm and thelayer-by-layer detection method into one fiamework in a probability sense.In the third part of the dissertation,we focus on numerically robust andlow-complexity V-BLAST detection algorithm over the spatially correlated channels.Using Householder transformation,and combining with the WY representation,afeasible V-BLAST detection algorithm based on the modified Householder QRD(M-H-QRD) is firstly proposed for the spatially correlated flat-fading channels.Theoretical analysis shows that the proposed M-H-QRD detection algorithm has robustnumerical property in contrast to the standard V-BLAST detection algorithm and thesorted QRD (S-QRD) detection algorithm over spatially correlated channel environments.Specially,compared to the S-QRD detection algorithm,the M-H-QRDdetection algorithm needs a smaller minimum word-length to reach the same value ofthe error floor for finite word length (FWL) precisions.Moreover,the proposedalgorithm can almost match the detection performance of the standard V-BLASTalgorithm in the moderate and low SNR region with much reduced computationalcomplexity.Then,for the practical purpose,an iterative detection algorithm based onthe M-H-QRD is proposed for V-BLAST system with the small number of receiver andtransmitter antennas.Simulation results show that the iterative scheme can achievemuch performance superiority over the standard V-BLAST algorithm with robustnumerical stability of the M-H-QRD.In the next part of the dissertation,we investigate the semi-blind channelestimation and decoding for the Alamouti space-time coded system over time-varyingfiat-fading channels,and two joint decoding and semi-blind channel estimationalgorithms are proposed.The first one,at the basis of channel estimation initializationby linear interpolation algorithm,employs the expectation-maximization (EM)algorithm to implement the joint Alamouti space-time code (STC)decoding and channelestimation in an iterative mode.The effects of the frame length of the data on theperformance of the algorithm are particully investigated.The other one combines theKalman filtering and the box-constrained ML (BCML) algorithm,in which the Kalmanfiltering is employed to tracking the time-varying channel,and then Alamouti STC isdecoded by BCML algorithm.Finally,we investigate the signal detection and channel tracking over the MIMOtime-varying frequency-selective channels and propose a complexity-controllablesequence detection method (i.e.Γ-H-MLSE method) with adaptive channel tracking.The proposed method has a certain reference value for extending the ideal of themaximum likelihood sequence estimation (MLSE) algorithm into multi-antennacommunication systems.