NATURE OF WASSERSTEIN METRIC ON WAVEFORM SIMILARITY EVALUATION AND EXAMPLE OF APPLICATION TO SEMBLANCE ANALYSIS

Abstract

 Similarity evaluation of seismic waveforms is one of the most important aspects in earthquake engineering and seismology. Recently, Wasserstein metric has been proposed as a new index for similarity evaluation of seismic waveforms, and although Wasserstein metric has been shown to be effective in a certain field, there has been no researches on its application as an index for waveform similarity evaluation in earthquake engineering. In this study, we applied Wasserstein metric to semblance analysis and compared it with the conventional method using the mean square error. The conventional method required a global search with high computational cost such as grid search due to the problem of multimodality, but when Wasserstein metric is used as the criterion function, it is shown that there is a certain degree of convergence to the globally optimal solution even with only local search with low computational cost.