Firstly, a reordered description is given for the linear minimum mean square error (LMMSE)-based iterative soft interference cancellation (ISIC) detection process for Mutipleinput multiple-output (MIMO) wireless communication systems, which is based on the equivalent channel matrix. Then the above reordered description is applied to compare the detection process for LMMSE-ISIC with that for the hard decision (HD)-based ordered successive interference cancellation (OSIC) scheme, to draw the conclusion that the former is the extension of the latter. Finally, the recursive scheme for HD-OSIC with reduced complexity and memory saving is extended to propose the recursive scheme for LMMSE-ISIC, where the required computations and memories are reduced by computing the filtering bias and the estimate from the Hermitian inverse matrix and the symbol estimate vector, and updating the Hermitian inverse matrix and the symbol estimate vector efficiently. Assume N transmitters and M (no less than N) receivers in the MIMO system. Compared to the existing low-complexity LMMSE-ISIC scheme, the proposed recursive LMMSE-ISIC scheme requires no more than 1/6 computations and no more than 1/5 memory units.