PROJECTS

This project develops TransECG, a Vision Transformer–based framework that analyzes ECG signals for re-identification risks while providing explainability. By applying attention mechanisms, TransECG pinpoints critical ECG components — such as the R-wave, QRS complex, and P–Q interval — that contribute to biometric identification tasks including gender, age, and participant ID. Evaluated on four real-world datasets, the model achieves high accuracy (≈89% across tasks) while revealing how ECG features encode personal identity. By combining predictive power with interpretability, TransECG supports the creation of privacy-conscious, secure, and trustworthy frameworks for ECG data sharing in healthcare and research.