Multi-input　multi-output　(MIMO)　systems　can　achieve　more　capacity　than　single　input　single　output　systems.　For　MIMO　systems,　maximum　likelihood　detection　is　the　optimum　detection　but　its　exponential　complexity　limits　its　application.　In　this　paper,　a　complex　lattice　reduction　algorithm　based　on　householder　transform　is　proposed　and　combined　with　MIMO　suboptimal　detection　algorithm.　Moreover　the　complexity　of　this　complex　lattice　reduction　algorithm　is　lower　than　that　of　real　lattice　reduction　algorithm.　Simulation　results　show　that　MIMO　suboptimal　detection　algorithm　based　on　this　complex　lattice　reduction　can　approach　the　optimum　performance　of　maximum　likelihood　detection　by　optimizing　channel　matrix　to　get　better　decision　domain.