An　empirical　mode　decomposition　(EMD)　based　stochastic　subspace　identification　(SSI)　method　is　presented　to　extract　the　modal　parameters　of　the　structure　induced　by　ambient　vibration.　The　measured　signals　are　first　decomposed　into　several　intrinsic　mode　functions　(IMF)　by　using　intermittency　frequency　and　each　IMF　represents　a　modal　response　where　a　natural　frequency　is　dominated.　The　decomposed　data　are　then　used　as　the　input　of　stochastic　subspace　identification　to　identify　the　modal　parameters.　A　case　study　of　a　real　bridge　ambient　vibration　test　has　demonstrated　that　the　proposed　method　can　overcome　the　difficulty　of　mode　mixing　and　reduce　the　noise　effect.　The　method　makes　the　stochastic　subspace　identification　more　clear　and　easier.