To suppress the cross terms interference in the Cohen class quadratic time-frequency distribution, a method based on empirical mode decomposition (EMD) and Choi-Williams distribution is proposed. In this method, the time domain signal is decomposed into intrinsic mode functions (IMFs) by EMD in frequency domain. It calculates Cohen class distribution after deleting the false components generated by EMD, and then the Cohen class distribution of original signal is reconstructed by superposing the results of IMFs to the original signal linearly. The simulation results show that the method is effective to suppress the cross terms of Cohen class Distribution, ensure Cohen class Distribution time-frequency concentration, and extract features of disturbance.