A　novel　modal　parameter　identification　method　of　ARMA　model　based　on　improved　Empirical　Mode　Decomposition　(IEMD)　subject　to　ambient　excitation　is　presented　in　this　paper.　It　is　able　to　partly　solve　the　problems　of　identifying　modal　parameters　in　ambient　excitation,　such　as　mode　mixing　and　false　mode　in　the　classic　EMD,　only　output　responses　and　the　difficulty　of　determining　the　order　of　ARMA　model.　At　first,　a　bandpass　filter　method　is　used　to　pre-process　the　measured　primary　signals,　and　the　sum　of　narrow-band　signals　is　obtained.　Then　a　series　of　Intrinsic　Mode　Functions　(IMFs)　are　separated　from　the　processed　signals　by　using　EMD.　The　real　IMF　is　determined　by　the　correlative　coefficients　between　the　separated　IMFs　and　the　primary　signals.　Finally,　the　Natural　Excitation　Technique　(NExT)　and　ARMA　(2,　2)　model　are　combined　to　identify　structural　modal　parameters　as　soon　as　the　real　IMF　is　obtained.　To　illustrate　its　effectiveness,　modal　parameters　of　a　7-storey　steel　frame　are　identified　with　the　proposed　method.　The　results　show　that　the　approach　proposed　can　extract　modal　parameters　effectively,　and　also　has　an　excellent　adaptability.　©　12　American　Scientific　Publishers.