A major problem encountered with both analog and digital accelerograms are distortions and shifts of the reference baseline. Accumulating experience indicates that direct integration of the ground acceleration data provided for seismic soil-structure interaction analysis often causes unrealistic drifts in the derived displacement. The drifts may have a significant effect on large-scale interaction analysis in which the displacement excitation is required as an input. The paper proposes a simple approach to remove long-period noise to acquire a realistic displacement-time series. In this approach, the acceleration data is firstly baseline-corrected in the time domain using the least-square curve fitting technique, and then using empirical mode decomposition (EMD) to extract signal trend to remove effect produced by long-period signal. Both simulated and field measured signal are employed in this study to demonstrate the feasibility of this approach.