Pattern Recognition Using Cluster Analysis Applied to Automated First Break Picking

Cluster analysis is a well-known statistical tool which is often applied when classification problems arise. The identification of first break signals may be interpreted as a specific classification or recognition problem. A "learning" classificator type is used to generate the initial database, which includes the manual choice of first breaks within some reference data and the calculation of probability densities describing the attribute space. Because the database may be updated during the automatic picking process, the system may adapt itself to changing signal and noise conditions.