Abstract: In order to solve the problem that time domain or frequency domain filtering is not effective in denoising for nonlinear frequency modulation signal, a novel adaptive time-frequency (TF) filtering method based on generalized S-transform is proposed. Firstly, the TF distribution spectrum of the signal is generated by applying generalized S-transform. Then, based on the TF distribution characteristics of the effective signal component, the TF pass region of the signal is identified, outside of which, the TF distribution of the noise is removed. In the next step, an adaptive TF filter is constructed using the TF concentration of the effective signal component, to suppress the noise within the effective signal component to obtain filtered TF distribution of the nonlinear modulation signal, which is then converted to time domain using inverse generalized S transform, to generate the filtered signal. Simulation results demonstrate that the proposed algorithm provides satisfactory performance in noise suppression and improves the signal-to-noise ratio.