Automatic Digitization of JMA Strong-Motion Seismograms Recorded on Smoked Paper Using Deep Learning

—A Case for the 1940 Kamui-Misaki-Oki Earthquake—

Abstract

Furumura et al. (2023) trained Convolutional Neural Network models to automatically digitize waveforms in the JMA strong-motion seismograms. To confirm the validity of the model, automatically digitized data of the smoked-paper strong-motion seismograms of the 1940 Kamui-Misaki-Oki Earthquake, which were not used for model training, were compared with manually digitized data. The comparisons showed that although some data correction was required, the automatic data were in good agreement with the manual data, and that this method can significantly improve the efficiency of analog record digitization.