In　recent　years,　with　the　continuous　progress　of　computer　deep　learning,　neural　network　and　other　technologies,　a　large　number　of　analysis　and　action　recognition　have　been　carried　out　for　various　types　of　motion　behaviors　through　the　establishment　of　relevant　models,　and　the　recognition　methods　of　human　behavior　characteristics　characterized　by　pictures　have　also　been　rapidly　developed.　In　this　paper,　a　three-dimensional　convolutional　neural　network　with　dual　resolution　is　established,　and　then　the　motion　technology　action　data　set　is　established　through　the　model,　Using　the　SSD　target　detection　algorithm,　the　feature　vectors　extracted　from　the　dual-resolution　network　are　fused　and　the　SVM　classification　experiment　is　carried　out.　The　final　experimental　results　show　that　the　algorithm　flow　designed　in　this　paper　is　based　on　the　motion　recognition　of　the　motion　technology　action　data　set　established　in　this　paper.　On　this　basis,　the　recognition　results　of　two　resolution　neural　networks　and　support　vector　machines　are　compared.　The　effectiveness　of　this　method　is　verified　by　the　simulation　of　the　constructed　sports　skill　behavior　samples.　©　2023　IEEE.