Emotion　processing,　playing　an　important　role　in　our　social　interactions,　is　a　sub-topic　of　social　cognition.　Significant　differences　in　emotion　perception　and　processing　have　been　demonstrated　between　schizophrenia　and　normal　people.　Therefore,　it　is　a　very　effective　strategy　to　use　the　emotional　stimulation　as　the　core　means　to　explore　the　difference　between　patients　and　normal　people,　and　then　to　develop　the　discriminative　model　for　patients　with　schizophrenia.　In　this　paper,　emotional　images　were　used　to　stimulate　the　two　groups　(schizophrenia　group　and　control　group),　and　the　electrophysiological　signals　during　the　experiment　were　recorded.　In　the　feature　extraction　phase,　the　time-domain　dynamics　and　the　asymmetry　of　the　hemisphere　were　considered　at　different　stimulation　stages.　Finally,　five　effective　machine　learning　methods　were　used　to　distinguish　between　schizophrenia　and　healthy　controls　under　positive　and　negative　emotional　stimuli,　respectively.　The　experimental　results　show　that　the　two　groups　of　event-related　electrophysiological　signals　obtained　by　negative　stimulation　can　be　better　distinguished　than　those　obtained　by　positive　stimulation.　And,　this　phenomenon　is　more　pronounced　in　the　time　window　of　first　second　after　the　stimulus　appears.　Meanwhile,　the　highest　average　F-score　with　10-fold　crossvalidation　strategy　can　reach　0.994　by　combining　both　support　vector　machine　classifier　and　grid　search　methods.