In　this　paper,　a　feature　subset　discernibility　hybrid　evaluation　method　based　on　the　discernibility　of　relative　distance　and　support　vector　machine　(DRD-SVM)　is　proposed　for　the　feature　selection　problem　of　the　upper　limb　rehabilitation　training　motion　of　Brunnstrom　4-5　stage　patients,　in　which　a　relative　distance　is　introduced　into　evaluating　the　discernibility　between　classes　considering　the　joint　effect　of　both　candidate　and　selected　features.　First,　a　feature　subset　search　strategy　is　used　to　search　a　set　of　candidate　feature　subsets.　Then　the　DRD　is　used　to　evaluate　the　candidate　feature　subsets,　the　best　subset　is　selected　as　a　new　selected　feature　subset,　and　the　feature　subset　with　the　best　performance　of　SVM　classification　is　selected　as　the　optimal　feature　subset.　Finally,　feature　selection　experiment　was　carried　out　on　upper　limb　routine　rehabilitation　training　samples　of　the　Brunnstrom　4-5　stage.　The　experimental　results　shows　that,　compared　with　the　F-score　method　and　the　DFS　one,　the　proposed　method　can　obtain　the　feature　subsets　with　higher　accuracy　and　smaller　feature　dimension,　which　improves　its　effectiveness　and　feasibility.　©　2018　IEEE.