Empirical mode decomposition (EMD), a new self-adaptive signal processing method, has been recently developed for nonlinear and non-stationary time series analysis. In this paper, EMD method is described and applied in time-frequency analysis. Aiming at the problems of intrinsic mode function (IMF) criterion in the EMD method, neural network (NN) prediction model and wavelet packet transform (WPT) technology are simultaneously introduced into the EMD method to improve the border effect and to enhance the ability of signal analysis, and thus a hybrid EMD-based time-frequency analysis strategy is proposed. The simulated time series are exploited to verify the effectiveness of the proposed hybrid model. Experimental results indicate that the hybrid strategy gives a quite satisfactory performance when both NN prediction model and WPT method are employed.