Cepstral time-frequency domain deconvolution based on improved generalized S-transform

Key words:  cepstral time-frequency domain  improved generalized S-transform  deconvolution  resolution

Abstract:

The key to time-frequency domain deconvolution is time-varying wavelet extraction and analysis of time-frequency spectrum. The traditional time-frequency domain deconvolution is influenced by Gabor transform time window fixation, and is based on the assumption that reflection coefficient satisfies white spectrum characteristic, leading to inaccuracy in deconvolution results. An improved generalized S-transform is introduced to obtain the time-frequency spectrum of non-stationary seismic records, which is then transformed into the cepstral time-frequency domain to achieve the time-frequency deconvolution. The conventional time-frequency domain deconvolution algorithm is improved by enhancing the stability of the algorithm, and by removing the white spectrum assumption in reflection coefficients. The comparison between the synthetic modeling results and the actual data suggests that this method can improve the resolution and can successfully model the attenuation of the seismic wave in seismic records.