Tensor-based Semi-Blind Channel Estimation Algorithm Research For MIMO Systems

Tensor decomposition is a new idea of low-rank decomposition of multidimensional matrix. In recent years, tensor decomposition has been widely used in the field of wireless communication and signal processing. System parameters and signals can be estimated by fitting the constructed three order tensor signal under the condition of uniqueness. In the wireless communication system, multidimensional signals can be constructed into the tensor model based on the system structure and the transmission mechanism, and the system parameters can be estimated based on tensor decomposition.Compared with the matrix arithmetic operation, tensor decomposition can make full use of the multi-dimensional information and separation each link of the channel matrix from the cascaded channel, then all the channel matrices are estimated simultaneously without error propagation, and then improve the channel estimation accuracy. In this thesis, tensor model is applied to the MIMO systems for channel estimation, then semi-blind channel estimation problems are studied, and the main work of the thesis are arranged as follows:1) For semi-blind channel estimation in MIMO systems, the received signals can be constructed into the multidimensional matrix by the number of received antenna and the length of code and the length of time frame. Under the condition of uniqueness decomposition, a novel estimation approach based on PARAFAC model is devised in this thesis. The simulation results show that the proposed method can loose the limitation on the number of antenna at the receiver compared with the traditional method based on pilot channel estimation, and the proposed method can obtain higher channel estimation precision with few pilot sequences.2) For semi-blind channel estimation in MIMO relay systems, the received signals can be constructed into the multidimensional matrix by the number of transmitted antenna, the number of received antenna, the length of time frame and the number of symbol block. Under the condition of uniqueness decomposition, a novelestimation approach based on two tensor models are devised in this thesis. The simulation results show that the proposed method has three advantages compared with the method based on pilot channel estimation. The first one is that it can loose the limitation on the number of antenna, and improving the system spectrum efficiency with few pilot sequences is the second one, and the third one is that each link of the channel matrix can be estimated simultaneously without error propagation,and then improve the channel estimation accuracy3) The regular alternating least-squares iterative algorithm is proposed to fit the PARAFAC model. Compared with the alternating least squares algorithm in relevant thesis, RALS iterative algorithm can eliminate ill-posed problem which come from the matrix pseudo inverse, and RALS has low algorithm complexity and faster convergence speed. This thesis analyzes the related parameters on the algorithm performance of the proposed iterative algorithm, which provide a basis for reasonable setting algorithm parameters. Two stages of iterative algorithm is proposed to fit PARAFAC and PARATUCK2 models for MIMO relay systems, and compared with the iteration algorithm in references thesis, the proposed iterative algorithm has low complexity.