This paper investigates application of ensemble empirical mode decompo sition (EEMD), which is a signal processing tool recently developed, to noise re duction of fatigue signals. A framework for noise reduction based on EEMD was pr oposed. The feasibility and effectiveness of the framework were verified by appl ying EEMD to the simulated signal, the experimental fatigue signal and the field measured strain time history. The results demonstrated that if the proper compu tational parameters are chosen, the EEMD method can effectively reduce the measu rement noise in the fatigue signal and significantly improve the accuracy of fat igue life prediction. Finally, rules were suggested for selection of computation al parameters for EEMD.