Decision-Theoretic Approach for Classification of Ricker Wavelets and Detection of Seismic Anomalies

Abstract

Decision-theoretic pattern recognition methods are applied to classifying Ricker wavelets and to detecting waveform anomalies in seismograms. The methods include Bayes decision rule and linear and quadratic classifications. Envelope and instantaneous frequency are extracted as the two features of a seismic trace used as input into the classification schemes. A modified fixed-increment training procedure is employed to solve the decision boundary. The classification schemes successfully distinguish zero-phase Ricker wavelets of different peak frequencies from each other and from random noise.