The early stage weak impulsive fault feature is so weak that it is always covered by environmental noise, which increases the fault diagnosis difficulty of rolling bearing. Aiming to this problem, a new diagnosis method based on adaptive optimal Morlet wavelet transform is proposed. Firstly, The core parameter of Morlet wavelet transform is calculated by particle swarm optimization (PSO) adaptively, which guarantees optimal Morlet based wavelet as well as wonderful band-pass filter performance; Secondly, in order to improve signal-to-noise ratio, optimal Morlet wavelet is used to filter incipient fault signal of rolling bearing; Finally, optimal Morlet wavelet filtered signal is analyzed by envelope spectrum, and the fault location of rolling bearing is extracted by contrasting the major frequency with the fault frequency of rolling bearing. The analysis results of simulated signal and measured signal show that the proposed method is able to extract the fault impulse signal.