In this study, we used the generalized S-transform method to improve the signal-to-noise ratio of ambient seismic noise data, and improved the efficiency of geophysical inversion. The S-transform method is a time-frequency representation of local spectral phase properties. A key feature of the S-transform method is that it can uniquely combine a frequency dependent resolution of the time-frequency space and absolutely reference local phase information. As such, the phase in a local spectrum setting can be defined, and this makes possible the production of many desirable wave characteristics. The scaling property of the Gaussian window is reminiscent of the scaling property of continuous wavelets, because one Fourier frequency wavelength is always equal to one standard deviation of the window. In this study, we introduce a variant of the original S-transform, which replaces f with λf^p. In this variant, one standard deviation of the Gaussian window contains λf^p wavelengths of the Fourier sinusoid at all frequencies. The generalized S-transform solves the defect in the analysis of non-stationary signals caused by the fixed form of the basic wavelet in the S-transform, and the time-frequency characteristics can better describe the non-stationary signals. We wrote a computer program for the generalized S-transform, and for ambient seismic noise we used data denoising. The results showed that the signal-to-noise ratio is greatly improved through the generalized S-transform denoising, and yields more accurate data for the inversion process. There have been few studies on the denoised processing of background noise data using the generalized S-transform, and the resolution of our processing results is better than that achieved by the ordinary S-transform.