In this paper, the authors made Ricker wavelet-based adaptive matching pursuit decomposition of seismic signals and discussed the method of time-frequency representation by matching wavelets. Since the time-frequency information of matching wavelet's Wigner-Ville distribution is limited, this paper proposes a time-frequency representation based on adjustable window Short Time Fourier Transform. Furthermore, by using the complex spectrum of matching wavelets, this paper introduces a new time-frequency representation, which not only has the same resolution and energy concentration properties as Wigner-Ville method, but also retains the most basic time-frequency characteristics of the original signal without cross term and window truncation effects. Compared with time-frequency characteristics of Short Time Fourier transform and S-transform, the matching pursuit time-frequency representation and instantaneous spectral parameters have higher resolution. Real data application also shows that the matching pursuit time-frequency analysis is very suitable for seismic signal with non-stationary characteristics.