Measuring dementia behaviours through depth sensors

Human Behavior Modelling aims to analyze human behaviors using visual or 3D data. In dementia, behavioral changes are strongly correlated with cognitive decline, making their monitoring essential. To address privacy concerns, especially important for vulnerable populations, this study uses depth maps instead of RGB images. We present a novel Human Pose Estimation (HPE) method using point cloud sequences derived from depth data and discuss domain gaps between benchmark and real-world datasets.