Target　material　recognition　is　of　great　significance　for　robot　grasping　and　manipulation.　Humans　can　easily　recognize　objects　through　a　combination　of　multiple　senses.　However,　it　is　difficult　to　distinguish　objects　with　the　same　shape　but　completely　different　internal　materials　for　a　vision-only　robot.　This　paper　proposes,　for　the　first　time　a　surface　material　classification　based　on　tapping　sound　features　to　make　up　for　the　lack　of　machine　vision.　The　algorithm　includes　audio　data　collection,　data　preprocessing　and　category　judgment,　etc.　To　evaluate　the　performance　of　the　algorithm,　we　performed　two　tasks:　distinguishing　the　category　of　objects　based　on　the　audio　and　judging　whether　the　two　audios　are　from　the　same　category　of　objects.　Experimental　results　show　that　the　accuracy　of　the　two　tasks　reaches　87%　and　98%,　respectively.　This　is　enough　to　prove　the　feasibility　of　the　algorithm　in　this　paper.　©　2021　IEEE.