Study On Signal Detection Algorithms For MIMO Systems

Multiple-input multiple-output (MIMO) technology has been considered as one of the core technologies in future wireless communication system because of the potential of achieving high channel capacity and spectrum efficiency without increasing the bandwidth or transmitted power. Signal detection is one of the key technologies and important steps in MIMO system. Signal detection algorithms for MIMO systems are studied in this thesis. The main contributions of the thesis are as follows:Classical ordered successive interference cancellation (SIC) algorithm suffers from error propagation and high complexity, so an improved SIC algorithm based on Maximum likelihood (ML) detection is proposed, in which signal detection is performed at two stages. ML detections for several layers are carried out firstly in order to reduce detection error, and redundancy of candidate sequences are selected to improve detection performance for next step. Sorted QR decomposition based SIC algorithm are performed in second step to reduced calculating complexity in second stage. By adjusting the numbers of layers in first stage and candidate sequences in the second stage, tradeoff between detection performance and calculating complexity can be obtained properly.QRD-M algorithm is a breadth-first searching scheme, which will meet high complexity when pursuing performance near ML detection, so depth-first QRD-M algorithm is proposed. Exhaustive search in the root node is performed firstly, and QRD-M search is carried out from the branch which has the smallest partial accumulated metrics (PAM) to then next branch serially with terminal condition. The PAM of the first found leaf node is as the terminal condition, when a new leaf node is found out which has smaller PAM, the condition would be updated, and when PAM of a branch is bigger than the terminal PAM, the search is over. The searching scope can be decreased greatly and the complexity is reduced with similar detection performance.Searching branches serially by QRD-M and depth-first QRD-M algorithm will lead to more time delay for the leaf node with the smallest PAM, so parallel QRD-M algorithm is proposed here. After extending the root node to its branches, QRD-M is performed parallel in each branch of the root node, which can improve the detection performance dramatically because of more nodes are searched in this algorithm. So another parallel QRD-M algorithm with partial sequences is proposed to trade off performance and complexity. By setting numbers of layers to perform ML detection and partial sequences to perform parallel QRD-M search, good trade-off can be get in the proposed algorithm.Group QRD-M algorithm is proposed here for pursuing detection performance like parallel QRD-M and lower complexity like depth-first QRD-M. ML detection of P layers are carried out, and L partial sequences are selected out with smaller PAM, then these partial sequences are divided into several groups, the possibility of containing the path with the smallest PAM in the first group will become great. Depth-first QRD-M detection is performed with each group in serial to search of the path near the performance of ML detection The numbers of ML layers, the partial sequences and groups can be set at different values to meet the different demands of practice.Chase algorithm can be seen as a unified frame of several detectors, by selecting different parameters in Chase algorithm, it can be deduced to SIC, parallel interference cancellation (PIC), tree search, and so on. Chase algorithm with sorted based on SNR maximum is proposed to deal with the phenomenon of error propagation. The method which can reduce the detection error is used to modify candidate list in Chase base on SNR maximum sorted. Chase algorithm can deduce to more detectors by changing the length of list and the type of sub-detector, which make Chase algorithm more universal.