Based on the selection of two important parameters k (the amplitude coefficient of white noise) and M (the number of ensemble members) in the process of applying EEMD to extracting mechanical faults, the law of how white noise with different amplitude coefficients affects the distribution uniformity of signal extreme as well as the decomposition precision is analyzed, and an adaptive parameter optimum method for EEMD based on the distribution uniformity of extreme points is proposed. According to the signal characteristics, k is adaptively obtained when the distribution uniformity of extreme points gets evenly distributed, and the M is obtained by calculating the expectation error. The feasibility and validity of the proposed method are verified with experimental results and industrial measurement analysis. This method shows obvious advantages when compared with traditional EEMD parameter-determining methods.