A new generalized least mean-square algorithm for processing non-stationary seismic data

Abstract

An adaptive deconvolution algorithm based upon a generalized expression of the Least Mean-Square (LMS) error technique is presented. The use of this process is recommended for reflection seismic data which contain timevarying reverberations. Filter coefficients are designed for each sample of the input trace using the proposed method. Convergence characteristics of the new algorithm, and its stability properties, are analyzed and compared to the simple LMS algorithm. Illustrations using synthetic seismic data are presented. Future possibilities of application to real shallow-water seismic data are found promising.