Self-supervised Human Activity Recognition

By Dr. Ong Wee Hong, Dr. Owais Ahmed Malik

---

SELF-SUPERVISED HUMAN ACTIVITY RECOGNITION

• Development of the self learning ability of an intelligent system to recognize everyday human activities without being trained with labelled data
• Investigate motion features – spatial, temporal, unsupervised: k-means, spectral, hierarchical, BIRCH, mean shift, DBSCAN, OPTICS, Gaussian Mixture Models, supervised activity models: HMM
• Technologies: computer vision, spatial and temporal features extraction, hybrid of unsupervised and supervised learning, human machine interaction
• Output: 7 publications