AI × AEC · CURATED INSIGHT
NCREE Evaluates AI Classifiers to Filter False Seismic Triggers
Structures · Smart operations
NCREE developed SVM classifiers using 6-second vertical signals to distinguish earthquake waves from vibrations, reducing false alerts.
NCREE reported monitoring stations suffer frequent false triggers; its Taipei building station logged non-seismic triggers over 40 times the seismic count. Testing SVM models on 6-second signals showed normalized Fourier amplitude spectra achieved over 95% precision and 85% recall.
Key takeaways
- Taipei station logged 7,255 non-seismic triggers versus 161 seismic events, risking alert errors.
- Normalized Fourier amplitude spectra performed best, though recall dropped across other stations.
Source：國家地震工程研究中心｜地震工程研究簡訊 Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human. Translation checked by AI.