Research On The Semi-blind Channel Estimation In MIMO-OFDM Wireless System

With the continuous development of communication technology,MIMO-OFDM technology has become one of the key technologies of the forth generation(4G) mobile communication systems. The combination of MIMO and OFDM technologies make the deficiency up between each other, and greatly improves system's high data rates and high channel capacity. The accurateness of channel estimation is crucial for high data rates and high reliability of MIMO-OFDM wireless system, and occupys a very important position in communication systems.The paper discusses the principle of MIMO-OFDM system, and based on the analysis of the fading characteristic of mobile wireless channel, summrises and compares the non-blind and blind channel estimation algorithms. Although simulation results show that the blind channel estimation performance is much better than non-blind channel estimation, there are still the problems of high computation cost, large of computing , estimated uncertainty and so on. Semi-blind estimation, which is an algorithm that adding few pilot tones in the blind channel estimation, could improve this situation to be better, and this method is a compromise of blind channel and non-blind channel algorithms. The paper mainly analyses the subspace-based semi-blind channel estimation algorithm. What the emphases analysed in the paper are the Projection Approximation Subspace Tracking (PAST) and Projection Approximation Subspace Tracking deflation (PASTd) methods for tracking signal subspace algorithm, and with both points of theirs, the paper gives an united Projection Approximation Subspace Tracking (UPAST) algorithm which uses signal subspace and noise subspace at the same time. The UPAST algorithm that the paper gives utilizes the contemporary eigenvalues and eigenvectors, which are given by the vectors of transmission signal with noise in the signal subspace and noise vectors in the noise subspace, to obtain eigenvalues and eigenvectors of the the transmitted signal in the signal subspace, and to ensure tracking signal subspace accurately. Theoretical analysis and simulation result show that,with the complexity of computing maintaining unchanged, the mean square error (MSE) and bit error rate (BER) decrease in some extent, and performance is better than them of the semi-blind channel estimations based on PAST and SVD subspace.