Due to the issues that low-frequency oscillation occurs frequently in interconnected grid it’s important to analyze the oscillation modes the fluctuation signal contains. In this work, a novel signal analysis method based on the combination of Independent Component Analysis (ICA) and Empirical Mode Decomposition (EMD) and Prony algorithm was presented. Through pretreating the observed power angle signals with filtering operation and its empirical mode decomposition to extract the intrinsic mode function (IMF), then processing the IMFs with whitening method, followed by using independent component analysis to get true IMFs, finally extract key oscillation modes from each IMF by Prony algorithm. Simulation results show that this method merged the advantages of both ICA and EMD which showed more competitiveness over the previous method, and help to improve the precision and accuracy of identification, to better meet the needs of practical application.