Research On Channel Estimation Algorithm For MIMO-OFDM Systems

Multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) technique is an effective solution which could implement high data rate and resist frequency-selective fading. Thus, the MIMO-OFDM technique is considered as one of the attractive techniques for future wideband wireless systems. Channel estimation for MIMO-OFDM systems is not only a challenging problem in wireless systems, but also an integral part of the receiver design. More importantly, the performance of channel estimation has a significant influence to the receiver, which accordingly makes the channel estimation become a hot topic in wireless communication areas.In this thesis, we systemically analyze the factors that may affect the performance of channel estimation in practical applications, and then, detailedly study on the major channel estimation algorithms for MIMO-OFDM systems. Concentrating on the frequency spectrum utilization, computational complexity, error floor and non-Gaussian noise problems in channel estimation for MIMO-OFDM systems, we propose some effective and achievable channel estimation algorithms.Based on the optimal pilot sequences, a least square (LS) channel estimation via angle domain is proposed. The phase-shifted orthogonal (PSO) pilot sequences is placed over multiple OFDM symbols, and the joint channel estimation is performed, which not only minimizes the channel estimation mean square error (MSE) but also reduces the requirement for pilot or enhances the frequency spectrum utilization. When the channel is spatially correlated, the LS algorithm, without using prior information, can not obtain good performance. So, depending on the approximately spatial independence of MIMO channels in angle domain, the most significant taps (MST) is retained by utilizing a suitable threshold, which improves the performance of the LS algorithm. The proposed algorithm satisfies the requirement for real-time process with low computational complexity and high frequency spectrum utilization.A channel estimation algorithm based on expectation maximum-maximum a posteriori (EM-MAP) is proposed. In order to reduce pilot overhead, the proposed algorithm splits each PSO pilot sequences of one OFDM symbol into multiple subsequences to different OFDM symbols. Then, statistical average is operated over multiple adjacent OFDM symbols to avoid the performance degradation due to the decreasing number of pilot subcarriers. After the estimated channel matrix is transformed from time domain to angle domain, the angle-domain filtering is used to improve the performance of the EM-MAP algorithm. Through the MSE analysis, it is indicated that the obvious decrease of pilot sequences can be obtained, stimultaneously, the performance of the EM-MAP algorithm is proportional to the number of OFDM symbols required in statistical average.An equivalent signal model (ESM) based channel estimation algorithm is presented. EM-MAP channel estimation algorithm will lead to the performance degradation at high signal noise rate (SNR) as a result of the error floor which is induced by the truncation of the channel impulse response (CIR) via EM algorithm. So, an accurate ESM which decomposes additive white Gaussian noise (AWGN) in traditional signal model (TSM) into two part indepent Gaussian noises was introduced. Based on the two-stage observation model and new complete data set introduced by this decomposition, a modified EM algorithm is derived. The performance of the EM and the EM-MAP algorithms is obviously improved at high SNR, by adjusting the proportion between these two Gaussian noises. Through the analysis of the ESM, the proposed algorithm retains CIR length and inhibits the error floor problem at high SNR, without increasing the complexity of design remarkably.A channel estimation algorithm based on the H-inf is proposed. When applied to MIMO-OFDM systems, objective function of traditional H-inf estimator could make the receiver design more complex and estimation less reliable, which makes us use the simplified objective function. Then, the EM and the space-alternating generalized EM (SAGE) iterative algorithms are used to convert MIMO channel estimation problem into SISO problems, which lower the calculational complexity of the H-inf algorithm. By using the ESM, next, the robustness of the EM/SAGE based H-inf algorithm is remarkably enhanced under non-Gaussian noise (NGN) condition. The proposed algorithm reaches convergence quickly, makes the real-time signal processing strongly and ensures the better peformance of estimation under NGN condition without using any prior information.