Seismic Signal Decomposition Using Blind Source Separation

Three-component seismic traces separated by BSS into P-like, S-like, and residual components

We investigate blind source separation of multicomponent seismic recordings to analyze overlapping wavefield components. Using the time-frequency and polarization structure of three-component observations, we explore how different arrivals and residual contributions can be distinguished.

Interdisciplinary collaboration

Related Publications

Yuta Amezawa, Tomohiko Nakamura, Takahiro Shiina, Satoru Fukayama, Jun Ogata, Hiroki Kuroda, and Takahiko Uchide, “Automatic detection and extraction of later phase in S coda using machine learning for crustal heterogeneity exploration,” in ACES (APEC Cooperation for Earthquake Science) International Workshop, Nov. 2025.
Yuta Amezawa, Tomohiko Nakamura, Satoru Fukayama, Takahiro Shiina, and Takahiko Uchide, “Automatic extraction and peak arrival estimation of later phase in S coda,” in International Joint Workshop on Slow-to-Fast Earthquakes 2024, Sept. 2024.