Research On Blind Equalization For MIMO System Baseed On Blind Source Separation

Multi input and multi output system is a communication system with multiple antennas in both the transmitter and the receiver,which is able to greatly improve the capacity of the system in a limited bandwidth.The actual MIMO system will be affected by both intersymbol interference and interuser interference,therefore,the source separation is required during the equalization.However,the traditional processing is based on the training sequence,which leads to an extra cost and reduces the bandwidth utilization.In this regard,it is an effective method to combine the blind equalization algorithm with the blind source separation algorithm,where the former is responsible for interference suppression,and the latter to achieve source separation.This paper mainly studies the combination of 3 blind source separation algorithms and different blind equalization algorithms,at the same time,some comparative analysis and improvements is given as well.These 3 blind source separation algorithms are the cross correlation algorithm,the multiuser kurtosis algorithm and the constrained fitting PDF algorithm.Under the same condition,the CFPA class has higher precision,then the MUK class,and finally the CC class.In terms of the convergence speed,the CFPA class is the fastest,the MUK class is the next,and the CC class is the slowest.However the complexity of the CFPA class algorithm is also the highest.Therefore,the focus of this study is to improve the precision of CC and MUK class algorithms,and to reduce the complexity of the CFPA with no or just a little loss of its performance.In view of the above problems,this paper proposes a series of improved algorithms.The main results obtained in this paper are as follows:1)By using the dynamic error feedback,the precision of the CC class and the MUK class algorithm is effectively improved: The steady-state error of the CC-MMA is reduced by 6~10d B,and that of the CC-MMA-DD is reduced by 1~2d B;and the steady-state error of the MUK-MMA algorithm and MUK-MMA-DD algorithm is reduced respectively by 9~11d B and 1~5d B;2)By using different cost functions at different stages,the computational complexity of CFPA class algorithm has been significantly reduced.At the same time,the loss of steady-state error of the CFPA-MMA algorithm is less than 1d B,and that of the CFPA-MMA-DD algorithm is less than 0.5d B.