Aiming　at　deficiency　in　identifying　modal　parameters　of　a　high　speed　rotor　system　using　modeling　approach,　a　new　method　for　modal　parameters　recognition　of　a　rotor　system　based　on　the　empirical　mode　decomposition　(EMD)　and　the　stochastic　subspace　identification　(SSI)　was　proposed　here.　Free　damped　response　components　of　the　rotor　system＇s　various　modes　were　obtained　with　the　narrow　band-pass　filtering　algorithm　from　the　rotor　system＇s　vibration　responses　under　operation　condition,　and　then　EMD　was　used　to　decompose　the　processed　signals　and　decrease　the　effects　of　mode　mixing　and　other　signal　components.　The　random　state　model　matrix　of　the　rotor　system　was　derived　from　a　Hankel　matrix　constructed　after　using　EMD　with　the　singular　value　decomposition　(SVD)　and　Kalman　state　filtering.　The　first　and　second　order　modal　frequencies　of　the　rotor　system　were　identified　with　eigenvalue　decomposition　of　the　random　state　model　matrix.　The　effectiveness　of　the　proposed　method　was　verified　with　simulation　analysis　and　actual　tests　on　a　rotor　system　rig.　©　2018,　Editorial　Office　of　Journal　of　Vibration　and　Shock.　All　right　reserved.