Abstract: Large-amplitude and long-period pulses are observed in velocity time histories of near-fault ground-motion records. The pulses in these records can pose severe ductility or strength demands to the near-fault structures and can subject them to higher collapse risks. Further research on the characteristics of pulses in near-fault ground motion is beneficial to deepen the understanding of the response of structures close to faults, and provide theoretical basis for the aseismic design. At present, methods related to strong motion processing and identification of near-fault pulses mainly focus on the single pulse in a record, so the multi-pulse characteristics of near-fault ground motions are less involved. Hence, a set of methods based on Hilbert-Huang transform （HHT） are proposed here to investigate the multi-pulse characteristics. Firstly, the raw near-fault record is corrected by the proposed HSA method, and then the ideal pulse signal can be extracted by the HHT method from the corrected record. According to the extracted pulse signal, the statistical relationships between pulse parameters and earthquake parameters are investigated. Finally, an effective strong motion duration is defined based on the pulse duration, which is verified by the nonlinear time history analysis of multi-storey buildings. The developed methods are particularly suitable for multi-pulse records. Stable peak ground displacement （PGD） and physically baseline offset time history can be obtained by the HSA method. Each velocity pulse in a record can be located in the time domain exactly and automatically by the HHT method. The proposed definition of strong motion duration for near-fault records can well characterize the intensity of multi-pulse records.