Research On Blind Channel Estimation For MIMO-OFDM System

For satisfying the demand of high-speed data transmission in the multi-media application of wireless communication, as two fundamental technologies of the future 4G communication, MIMO and OFDM have been concerned extensively. The combination of MIMO and OFDM have taken advantage of spatial diversity, frequency diversity and temporal diversity so as to greatly improve the channel capacity and transmission rate and effectively resist to multi-path fading, interference and noise. In addition, both the coherent demodulation of the receiver and the decoding of space-time coding should require the perfect knowledge of channel state information. Consequently, the channel estimation is extremely significant to the design of the receiver.Recently, the approach of the channel estimation of MIMO-OFDM system could be roughly classified as two classes. The first channel estimation method based on pilot or training sequences is non-blind. Because of its requiring a large amount of training sequences and wasting partial frequency resources, such method could not meet the demand of the future high-speed digital communication. The other blind estimation method based on few pilot or training sequences mainly uses the second-order statistics of receiving dates such as noise or signal subspace decomposition algorithm, comparing with non-blind estimation algorithm, it has excellent utilization ratio of frequency resources. But the disadvantages such as high computational complexity, slow convergence and bad robustness have held back the practical application of such method. So how to solve these problems have been the hot researching issues.The blind channel estimation of MIMO-OFDM is studied in this dissertation. Firstly, the primary principle, channel model and the blind and non-blind channel estimation algorithms for MIMO-OFDM have been introduced. Because of the superior performance against the structured noise, a novel approach for MIMO-OFDM blind channel estimation based on oblique projection has been proposed in the dissertation. When identifying the system, the receiving signal space is divided into"past subspace","present subspace"and"future subspace". Project obliquely"present subspace"onto"past subspace"and"future subspace"respectively and then the impact on the receiving system of channel multi-path effect can be decreased effectively. To improve the algorithm's ability of anti-noise, Singular value decomposition (SVD) method is applied.Finally, By comparison of estimation performance between oblique projection and noise subspace through the theoretical analysis and simulation, it's found that the oblique projection has lower computational complexity with slightly sacrificing the estimation accuracy.