Because the precision of modal parameter identification based on Hilbe rtHuang transform is affected by the mode mixture of empirical mode decomposit i on (EMD) and average times of random decrement technology (RDT), a novel modal p arameter identification method using stratified sampling and bandwidth restricte d EMD is proposed for short data sequences. A bandwidth restricting signal was e mployed in the procedure of EMD to restrain the mode mixture and to increase the frequency precision of EMD. In order to obtain more subdata sequences of RDT, stratified sampling, which took the place of random sampling and had the ability to improve precision of sampling and total average times, was employed to extra ct free decrement response signal. The method based on fitting difference and sa mple size was proposed to determine the layer’s weight of stratified sampling. The simulation and the application demonstrated that the proposed method is effe ctive and capable of improving precision of HHTbased modal parameter identific ation for short data sequences.