In the signal detection algorithms for multiple input multiple output (MIMO) system,maximum likelihood (ML) detection offers the optimal performance while it encounters difficulties in practical system because of its computational complexity,which is increased exponentially with the number of antennas and modulation order.Minimum mean square error-ordered successive interference cancellation (MMSE-OSIC) algorithm has low complexity,but comparing with the optimal detection,its detection performance has a substantial margin due to error propagation in the progress of iterative detection.For dealing with the above algorithms' shortcoming,a new detection algorithm with low-complexity and near-optimal-performance is proposed.The algorithm makes readjustments of decoding order in MMSE-OSIC algorithm,which begins with the determination of the weakest emission signal layer by comparing the maximum norm of row vectors,then makes exhaustive search on this transmitted signal;on the premise of the correct detection of weakest signal as possible,MMSE-OSIC algorithm is used for the detection of the remaining signal layers.Theoretic analysis and simulation results show that the algorithm can effectively strain error propagation in the progress of iterative detection and nearly reach optimal performance at low computational complexity.An appropriate trade-off between detection performance and computation complexity is obtained by this improved algorithm.