BROWSE
Detail
Conference Paper
Earthquake False Alarm Detection Model Augmented with Sparse Profile Analysis
Cited 1 time in 
Share 
Share
- Authors
- Jungeun Yoon, Aekyeung Moon, Seung Woo Son
- Issue Date
- 2025-08
- Citation
- International Geoscience and Remote Sensing Symposium (IGARSS) 2025, pp.6078-6082
- Language
- English
- Type
- Conference Paper
- Abstract
- Machine learning (ML) and deep learning (DL) have recently made significant advancements in earthquake detection, with models increasingly focused on minimizing false alarms and ensuring real-time responsiveness. Sparse profile analysis is utilized in this paper to enhance the accuracy and response speed of earthquake alert systems while minimizing false alarms. The proposed approach employs the discrete cosine transform (DCT) to convert inertial measurement unit (IMU) sensor data into the frequency domain and extract the sparsity representation of a signal within this domain. These extracted sparsity features, alongside statistical properties from the original IMU signals, are used to train several ML and DL models for classifying false alarms and actual seismic events. The experimental results demonstrated that the random forest model, augmented with features extracted through sparsity representation in DCT, achieved superior performance with an accuracy of 92.38%, a recall of 88.83%, a false positive rate (FPR) of 3.06%, and a detection time of 0.017 seconds. These results indicate that the models enhanced with DCT-based sparsity features significantly lower false alarms, achieving a quick detection time of under one second, emphasizing their suitability for reliable earthquake early warning systems.
- KSP Keywords
- Detection model, Discrete cosine Transform, Early Warning System(EWS), Earthquake early warning, False Positive Rate, Frequency domain(FD), Machine learning (ml), Profile analysis, Random Forests(RF), Random forest model, Response Speed