This　paper　studies　the　parallel　feature　level　fusion　algorithm　based　on　multiple　dimension　reduction.　In　view　of　the　traditional　serial　and　parallel　feature　fusion　method　shortcomings,　this　paper　proposes　a　dimensionality　reduction　method　for　the　feature　vector　using　PCA　(Principal　Component　Analysis)　method　before　fusing　the　feature　vector.　In　order　to　solve　the　high-dimensional　problem　after　feature　fusion,　this　paper　puts　forward　a　kind　of　generalized　K-L　transformation　based　on　the　unitary　space　to　compress　the　dimension　of　fusion　feature　vector　and　remove　redundant　data.　©　2016　ACM.