Abstract: When using successive interference cancellation (SIC) in MIMO detection, overall performance is determined by the first stages being decoded due to error propagation. To solve this problem, an improved SIC detection algorithm was proposed. For the algorithm undergoing SIC detection, an exhaustive search of initial P level signals was conducted, and accumulated metrics were calculated and arranged in order. From these were selected L signals with the smallest metrics. Then SIC algorithms were used to detect ongoing levels. Exhaustively searching initial P levels reduced frequency of coding errors and consequent error propagation and improved detection ability of the system. An appropriate trade-off between detection performance and computation complexity was obtained by adjusting P and L values.