Deep clustering of traffic signals using a single seismic station

In this study, we present an unsupervised deep clustering framework based on Deep Embedded Clustering (DEC) to identify traffic-induced signal segments from high-frequency seismic noise recorded by a single seismic station. This method enables cost-effective event extraction from ambient noise without labeled data or dense sensors, offering a alternative for urban traffic monitoring, particularly in areas with limited data or few monitoring devices. The presented code and dataset are part of our study on using DEC to analyze traffic-induced seismic signals at urban scales.