MIMO Channel Estimation Based On Reconstructed Observed Signal Sequence

Because of highly space resource using efficiency and highly spectrum using efficiency and improving channel capacity greatly without increasing channel bandwidth, Multiple-Input Multiple-Output system is chosen as one of the advantaged techniques for the future mobile communication system, which has been the topic of many researches.In general, using space time code can improve MIMO system capacity. However, the precondition of using space time code is that MIMO channel information can be estimated correctly. Only with the correct estimated channel information, MIMO system can get the ideal channel capacity. Therefore, how to get the MIMO channel information correctly becomes a crucial problem.Channel estimation, one of crucial techniques in MIMO system, is investigated in this paper. At first, from the traditional channel estimation strategy using training sequence, a novel channel estimation method, which is based on reconstructed observed signal sequence is proposed. Based on redundant information used by the novel method, two new observed sequences are proposed: one is accumulative observed noisy signal sequence (AONSS); the other is concatenated observed noisy signal sequence (CONSS). Under the assumption that channel is black and flat fading, firstly, maximum likelihood (ML) channel estimation, least square (LS) channel estimation and linear minimum mean square error (LMMSE) channel estimation, which are all based on AONSS and CONSS, are derived. Then, based on AONSS and CONSS, the generalized random Cramer-Rao lower bounds, which include traditional Cramer-Rao lower bounds as special cases, are also derived. Comparing to traditional ML, LS and LMMSE estimation algorithms, the performances of proposed ML, LS and LMMSE based on AONSS and CONSS are better. However, the precondition of using AONSS is that every transmitted package's training sequence is the same, using CONSS doesn't have such precondition. Comparing to AONSS, the shortage of CONSS is this method can increase the algorithmic complexity.Furthermore, we apply the ML channel estimator based on AONSS and CONSS, and the optimal training sequences for block fading channels to continuous flat fading channels and analyze the estimation error. We show that the channel estimation error for continuous fading channels is caused by noise as well as the temporal variation of channel. For the more observed training sequence signals, novel ML estimators based on AONSS and CONSS have less estimation error due to noise than that of traditional ML estimator. However, for the more observed training sequence signals, estimation error due to temporal variation of channel is increased. Furthermore, traditional ML estimator can be served as a special case of ML estimator based on AONSS and CONSS without redundancy. Simulation results show that under the continuous fading channel, the proposed reconstructed observed signal sequence method is viable for channel estimation, which gives another method to research the MIMO channel estimation based on reconstructed observed signal sequence.