The Research On MIMO Detection Methods Based On Matrix Preconditioning Techniques

Compared with the conventional single input single output system,the multiple input multiput output(MIMO)system brings in the space mulplexing,and it can provide array gain,diversity gain and mulplexity gain,so that it can exihibit better realibility and efficiency which can allievate the scarce spectrum resources.With the higher and higher command,the number of antennas becomes large nowadays.As an important part in the MIMO system,MIMO detectors have one critical contradiction needed to be resolved,i.e.,on one hand,if the detector has low complexity,then it can not provide full receive diversity in the MIMO system;on the other hand,the performance-optimal detector will cost large flops in detecting the signal,and the complexity grows exponentially with the increasing number of antennas.With the increasing of the number of antennas in the MIMO system,the system structure becomes more and more complex,and the contradiction becomes more serious.It has been shown that this contradiction is mostly affected by the characteristics of the channel matrix,e.g.,bad orthogonality and large condition number will enhance the contradiction between the complexity and the error probability performance.In this thesis,the matrix preconditioning techniques are used to improve the characteristic of the channel matrix,so that an equivalent channel matrix which is more suitable for detection can be shown,and the contradiction in the MIMO detection will be decreased.Moreover,the object of this thesis is the linear detectors and the ordered successive interference cancellation(OSIC)detectors,and the aim is to make up the flaw that low complexity detectors don't have small error probability,i.e.,the error probability performance of the low complexity is improved at most degree.The content is as follows.First,this thesis utilizes the permutation matrix to precondition the channel matrix,so that in the preconditioned equavilent channel model,the error propagation in the SIC(Successive Interference Cancellation)detector will be decreased and the error probability performance of the SIC detector will be improved(the OSIC detector is the SIC detector under the permutation matrix preconditioning).In the process of designing the permutation matrix,the global search algorithm is replaced by the greedy algorithm,and more information(channel matrix,the received signal in the receiver side and the modulation in the transmit side)in the receiver side is utilized,so that it can explicitly figure out the factor affecting the OSIC detector.Compared with the previous designing process of the permutation matrix,the designing methods in this thesis can provide better performance in reducing the error propagation in the SIC detector,so that it can provide better error probability performance in the SIC detector.Moreover,our algorithm is greedy,then it has low complexity,and this is very efficient when the number of the antennas in the MIMO system becomes large.Finally,the simulation results show that the SIC detector based on this matrix can provide efficient error probability performance and complexity.Secondly,this thesis analyzes the influence caused by the unimodular matrix in the MIMO linear detector(LD),and designs efficient unimodular matrix to improve the performance of the LD.In the designing process,the aim is to use the unimodular matrix to change the most important factor affecting the error probability performance of the LD,i.e.,modified Euclidean distance(MED).As the MED is directly related with the error probability performance of the LD baded on the unimodular matrix,so the error probability of the LDs will be improved if efficient unimodular matrix is designed to change the MED.Moreover,this thesis studies another important problem in the LD based on unimodular matrix,i.e.,the feasible set problem.In the previous detection process in the LD based on the unimodular matrix,the feasible set is approximately regarded as the integer field without constraint,and this approximation will not be exact when the constellation size is small,then the error performance of the LD based on unimodular matrix will become worse.In this thesis,the feasible set of the LD based on unimodular matrix is analyzed,and it is shown that it is the integer set in the polyhedron.Then if the symbol is detected in this feasible set,the error probability performance of the detector is further improved.Finally,the simulation results figure that both the proposed schemes can show better error probability performance than the previous schemes.Finally,this thesis analyzes the influence in the OSIC detector caused by the unimodular matrix,and gives two efficient unimodular matrix designing processes to improve the error probability performance of the OSIC detector.The first designing process takes the iterative process of the OSIC detector into account,and it targets to improve the signal to interference plus noise ratio of the detected symbol in each iteration.As the signal to interference plus noise ratio is an important factor affecting the error probability of the detectors,then based on this designing process,it can not only provide better error probability performance of the detected symbol in each iteration,but also the error propatation will be reduced,so that error probability performance of the OSIC detector will be improved.In the second designing process,a way similar to the designing process of the permutation matrix in the OSIC detector is proposed,i.e.,the designing process utilizes many informations,including: channel matrix,received signal and the constellation set,so that the error probability of the detected symbol in each iteration of the OSIC detector can be calculated.Then using the unimodular matrix to reduce the error probability of the detected symbol in each iteration of the OSIC detector,the error propagation in the OSIC detector will be decreased and the error probability performance will be improved.Finally,the simulation results figure that the OSIC detector based on both these two schemes can show better error probability performance than the previous schemes.