The Research On Channel Estimation Algorithms In MIMO-OFDM Wireless Communications Systems

One main target of next generation wireless mobile communication is to obtainmore efficiency of spectrum. In wireless communication with the limitation ofbandwidth and power, MIMO is a potential technique to achieve high efficiency ofspectrum, and considering OFDM technology has the property of high tolerance tomulti-path fading and low complexity to receiver desigen, MIMO-OFDM techniquehas been widely regarded as one of the most promising techniques for Beyond 3rdGeneration(B3G) wireless communication, which is efficient for widebandmultimedia transmission because of its high ability of anti-fading, high channelcapacity and high bit-rate data. As a key technique for wireless mobilecommunication, channel estimation is required for coherent detection and space-timedecoding in MIMO-OFDM systems. So channel estimation for MIMO-OFDM systemis the main research content in this dissertation.The wireless transmission channel with high rates has propertyof sparsitywhichcan be used to improve the performance of channel estimation. For time-invariantsparse MIMO-OFDM channel, a channel estimation algorithm based on generalizedAkaike information criterion(GAIC) is proposed, the channel order and the timedelays of each path can be identified by the algorithm, then the tap ofnon-dissemination path in the time-domain impuse reponse which is estimated byleast square (LS) algorithm is set tozero. Bythis method, performance loss caused byAWGN can be reduced and the channel estimation precision can be improved,meanwhile, as the number of channel path decreases, the proposed algorithmperforms better.Wireless MIMO-OFDM communication channel has property of fasttime-varying due to the high mobility of terminal. The time-selective fading channeldestroys the orthogonality among different subcarriers and cause Inter-CarrierInterference (ICI), then, sytem performance will decrease. A discrete prolatespheroidal sequence based expansion model is established for time-vaying channel,and an iterative channel estimation method that isn't depended on the statisticalinformation of CSI is presented. To cancel the interference from the unkowninformation, the equalized symbol is reused in channel estimation. Then the fasttime-varying MIMO-OFDM channel is estimated using pilot sequence by the abovemethod.Pilot-symbol assisted modulation (PSAM) scheme is usually adopted by channelestimation, but this scheme requires extra system bandwidth. To improve efficiencyof bandwidth, a superimposed training (ST) based channel estimation method is researched. It means that a periodic training sequence with low power issuperimposed to the information sequence at transmitter, with the mean of receiceddata, channel parameters can be estimated without consuming any extra systembandwidth. The unknown information sequence can be interference to channelestimation due to replacing the statistical mean with arithmetic mean. Based on thisan iterative superimposed training (IST) method is presented. The method uses sumof equalized information sequence and superimposed training as new trainingsequence, which is feedback to channel estimator to estimate channel stateinformation (CSI) again. The channel estimation performance can be improved bythefeedback iterative method. First time-invariant SISO-OFDM channel is estimated bythis method, then the method is extended to fast time-varying MIMO-OFDM channelestimation based on discrete prolate spheroidal sequences based expansion model.There exist colored interference and channel spatial correlation in multi-userMIMO-OFDM system, which result in performance degradation of channelestimation. To improve channel estimation precison, a space-frequency statisticalprecoding scheme is proposed in the dissertation. The precoding scheme is designedat transmitter by utilizing water-filling approach based on the minimum criteria ofchannel estimation mean square error (MSE) under the condition of total powerconstraint. It only requires statistical information of channel correlations and coloredinterference, which need not be updating frequently, and can be easily acquired.Theoretical analysis shows that the optimal transmission directions are determined bythe eigen-decompositions of channel covariance matrices and interference covariancematrices.