Frequency　modulated　empirical　mode　decomposition　(FM-EMD)　method　is　proposed.　The　mode　mixing　problem　associated　with　the　original　EMD　method　when　processing　nonlinear　signals　with　closely　spaced　frequencies　is　solved.　The　physical　meaning　of　the　identified　intrinsic　mode　functions　(IMFs)　is　improved.　Assisted　by　the　Hilbert-Huang　transform,　the　proposed　method　can　identify　the　instantaneous　parameters　of　the　weak　nonlinear　system　with　closely　spaced　frequencies.　The　effectiveness　of　the　proposed　method　is　verified　by　numerical　simulations.　With　the　proposed　method,　nonlinearity　is　identified　in　the　aeroelastic　vibration　of　the　two　DOF　bridge　deck　sectional　model　of　Humber　Bridge.