Mild　cognitive　impairment　(MCI)　is　highly　likely　to　convert　to　Alzheimer＇s　disease　(AD).　The　main　approach　to　identifying　MCI　is　using　a　functional　connection　network　(FCN).　Traditional　FCN　is　used　to　study　the　correlation　between　two　brain　regions,　but　it　lacks　deeper　brain　interaction　information.　Neuroscientists　found　the　internal　functional　activity　pattern　in　the　human　brain　is　characterized　by　sparse,　modular,　and　overlapping　structures,　and　the　FCN　is　restricted　by　the　brain　structural　connection　network　(SCN).　They　can　improve　the　estimation　accuracy　of　FCN.　Therefore,　this　article　first　constructs　low　order　FCN　(LFCN)　based　on　brain　sparse,　modular,　and　overlapping　activity　patterns.　Then,　new　high-order　FCN　(HFCN)　is　proposed　based　on　the　restrictive　relationship　between　SCN　and　FCN.　To　combine　high　robustness　of　LFCN　with　high　sensitivity　of　HFCN,　a　new　combination　strategy　of　LFCN　and　HFCN　is　proposed.　It　integrates　the　idea　of　brain　modular　and　overlapping　with　the　restricted　relationship　between　SCN　and　FCN.　Finally,　the　experimental　results　show　that　in　early　MCI　(EMCI)　recognition　the　best　classification　performance　is　acquired　with　an　accuracy　of　91.42%,　which　is　better　than　similar　methods.　This　method　will　be　instrumental　in　the　early　recognition　of　clinical　MCI.