A method of time-frequency analysis for quick access recorder (QAR) data is proposed, which combined the ensemble empirical mode decomposition (EEMD) and Hilbert transform (HT). Firstly, the paper briefly introduced the basic principles and its solution steps of EEMD. After evaluated the degree of stationary of the actual QAR data, the sequence which has small mutation is chosen to do in-depth analysis. The marginal spectrum reveals the EEMD could effectively eliminate mode mixing. And the Hilbert spectrum and the degree of stationarity curve clearly show the small mutation, which confirms the method proposed is correct and useful.