time-series analysis

My research focuses on the interdisciplinary application of time-series analysis to decode complex behaviors in both human activity and animal ecology. By leveraging advanced sensing technologies, I develop robust frameworks for Human Activity Recognition and automated behavior identification. My work specifically addresses critical technical bottlenecks in real-world deployments: label scarcity, model pre-training, and system interpretability.

Currently, my research strategy is built upon three pillars: progressing from motif detection for pattern discovery, to active/contrastive learning for reducing expert labeling costs, and finally employing synthetic IMUs to bridge data gaps. These innovations provide a scalable toolkit for various applications, including proactive risk prediction in manufacturing, intelligent monitoring in elderly care, and non-invasive behavioral tracking of wildlife.