The　tracking　and　extraction　the　hand　pose　is　the　previous　work　of　gesture　analysis.　I　n　this　paper,　a　mixed　Gauss　model　based　level　set　contour　extraction　algorithm　is　presented.　The　algorithm　combined　with　the　mean　shift　algorithm　to　achieve　the　extraction　of　hand　contour　dynamically　in　a　video　sequence.　This　method　first　get　the　initial　contour　by　color　information,　and　build　Gaussian　model　of　the　inside　and　outside　of　the　outline　of　the　image　features　by　using　Gaussian　mixture　model.　Depending　on　the　characteristics　of　the　Gaussian　distribution　and　combining　with　active　contour　algorithm　for　image　segmentation,　then　obtaining　hand　movement　position　by　mean　shift　algorithm,　we　achieved　gesture　tracking.　The　test　results　show　that　compared　to　pixel-based　active　contour　algorithm,　this　method　has　greater　stability,　which　can　get　more　accurate　extraction　region　and　achieve　accurate　contour　extraction　and　gesture　tracking.