Frequency Domain Seismic Blind Deconvolution Based on ICA

Independent components analysis (ICA) is combined with shot time Fourier transform(STFT) to improve the traditional frequency domain seismic deconvolution in this paper. Neglecting noise, the seismic record is changed from time domain to frequency domain with STFT in order to transform the common seismic model to the basic ICA model. By applying FastICA algorithm, reflectivity series and the seismic wavelet can be produced in frequency domain and changed back to the time domain subsequently. The model and real seismic data numerical examples all show the algorithm valid. The advantage of this new method is to inverse blindly the wavelet and the reflectivity effectively with no assumption of Guassality and whiten noise to reflectivity, and no assumption of minimum phase to seismic wavelet. The algorithm refered here propose a new way of seismic deconvolution and worth doing more researches.