Aiming　at　the　problem　of　lowering　the　success　rate　of　grasping　due　to　the　different　characteristics　of　the　grasped　object,　such　as　quality,　soft　and　hard,　in　the　process　of　robotic　garbage　sorting,　this　paper　proposes　an　MTV　tactile　character　recognition　model　based　on　the　attention　mechanism.　The　model　is　trained　using　five　tactile　capture　data　sets　for　tactile　traits,　and　the　recognition　rate　of　the　object＇s　hardness　and　hardness　traits　can　reach　more　than　90　degrees.　Experimental　results　show　that　the　model　can　effectively　identify　tactile　traits　with　a　high　recognition　rate.　Has　good　generalization　ability.