Abstract: A low complexity sub-optimal soft-detection algorithm for MIMO system is presented. Usually, the channel decoding needs every bit’s soft-information outputted from the MIMO detection in the iteration decoding for the coded MIMO system, but the calculation of the soft-information causes great complexity. The estimation of transmitted vector is obtained by using the hard decision of sphere decoding for the first iteration and using the sign of the priori information of the coded bits in other iterations in this paper. And then the soft-information is calculated for every bit of MIMO detection based on the estimation of transmitted vector. Therefore, the proposed algorithm avoids the conventional detection algorithm which is the exhaustive search, and decreases the computing complexity. The analysis and the simulation show that this algorithm reduces the complexity greatly at the cost of limited deterioration on performance. As for a frame data, the simulation’s time of the proposed algorithm is less than 1/20 of that of the conventional detection algorithm in the same setting and the complexity of the algorithm is constant on the whole for different modulations, it gets the better tradeoff between the performance and the complexity.