In　recent　years,　the　smart　home　has　been　a　breakthrough.　Mobile　speech　recognition　system　plays　an　important　role　in　smart　homes,　but　the　noise　generated　by　the　complex　and　changeable　home　environment　has　become　one　of　the　main　obstacles　to　the　use　of　speech　recognition　technology.　In　the　process　of　voice　interaction　(such　as　speech　recognition　and　control　functions　in　the　smart　speaker),　it　will　also　be　affected　by　the　noise　in　the　home　environment.　Through　the　scene　application　of　speech　enhancement　algorithm,　the　noise　signal　received　by　the　system　is　suppressed,　speech　intelligibility　is　increased,　and　the　speech　naturalness　is　improved,　so　as　to　improve　the　accuracy　of　semantic　conversion　and　speech　recognition.　©　2021,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.