The Research On Decoding Algorithms Constrained By Condition Number For Wireless MIMO System

MIMO wireless technology is a great breakthrough in modern wireless communications.Comparing with the traditional SISO system, MIMO system can significantly improve the channelcapacity and spectrum utilization rate without increasing the bandwidth and transmission power.However, MIMO wireless technology also brings some problems with this tremendous progress,and the detection algorithm is the important one of these problems.This thesis first introduces MIMO system and wireless channel model briefly. Then, on thebasis of these introductions, it makes an in-depth research on detection algorithm in MIMO system.Firstly, the thesis analyzes several heuristic detection algorithms in MIMO system, includingmaximum likelihood (ML) algorithm, linear detection algorithm and nonlinear detection algorithm,and compares their bit error rate (BER) performance through simulation results.Secondly, the thesis introduces the conception and properties of condition number of matrix,and the application of condition number in numerical analysis and wireless communication system.The thesis proves that the MIMO channel capacity is highly influenced by the channel matrixcondition number at high signal noise ratio (SNR), and the condition number of matrix also hassignificant impact on the performance of detection algorithms. The performance of detectionalgorithm would be better if the condition number of channel matrix become lower.Finally, it presents a new detection algorithm, named condition number threshold detectionalgorithm, for improving the performance of detection and reducing computational complexity ofdetection. The proposed algorithm combines ML detection and zero force (ZF) detection based onLenstra Lenstra Lovasz (LLL) algorithm which can reduce the condition number of channel matrixand improve the performance of ZF detection. When condition number of channel matrix is below aprescribed threshold, the proposed algorithm chooses ZF detection base on LLL algorithm forMIMO system, and ML detection otherwise. Thus this proposed detection can not only improve theperformance of detection but also reduce its computational complexity.