A　new　method　to　voiced/unvoiced/silence　of　speech　classification　using　Support　Vector　Machine　(SVM)　is　proposed.　This　classifier　can　effectively　classify　speech　frames　into　voiced　frame,　unvoiced　frame　and　silence　frame　under　various　levels　of　signal　noise　ratio.　Firstly,　in　high　SNR,　the　VU/S　classification　is　done　by　using　the　four　difference　characteristic　parameters　used　in　G.729B　VAD　as　SVM＇s　input　features.　The　comparison　of　experiment　results　shows　that　the　proposed　method　outperforms　other　traditional　methods　(G.729B　VAD　and　BP　network),　which　shows　the　SVM＇s　classification　method　is　available.　And　the　performance　of　SVM　for　different　kernel　functions　in　the　experiment　was　analyzed　and　discussed　as　well.　Secondly,　the　paper　also　discusses　the　extraction　of　the　statistical　features　which　is　immune　to　the　background　noise　and　the　adaptive　estimation　method　for　the　time-varying　background　noise　in　low　SNR,　which　are　analyzed　by　applying　a　statistical　model.　Lastly,　the　comparison　experiment　results　in　various　noise　environments　under　varying　levels　of　SNR　are　given.　According　to　the　simulation　results,　the　proposed　method　shows　significantly　better　classification　performances　than　the　other　traditional　methods　in　middle　and　low　SNR　cases.