Abstract: An adaptive Doppler distortion correction method is proposed. The two kinematics parameters, the sound source moving speed v and the the distance, x, between the microphone and the sound source at the initial moment, are used as the optimization variables, and the maximized frequency domain statistical index of the resampled signal is the optimization objective. The values of v and x are estimated through parameter optimization, and the distortion correction is performed through amplitude restoration and time-domain interpolation fitting. The simulation analysis results show that the four statistical indicators including the spectral kurtosis, spectral skewness, spectral impulse factor and spectral crest factor can accurately identify the kinematic parameters. The spectral kurtosis has the best anti-noise ability, and the critical signal-to-noise ratio reaches -3.1 dB. The experimental analysis results show that after the Doppler distorted sound signal of the faulty train bearing is corrected, the fault frequency component and higher harmonic components of the envelope spectrum are clear and accurate, indicating that the Doppler distortion is well corrected. This method can realize the comprehensive correction of the time-frequency structure of the Doppler distorted signal based on the signal itself.