The thesis focuses on developing contactless sensing solutions for everyday use, specifically targeting human vital signs and user-object interactions by leveraging radar vibrometry. Current sensing technologies for personalized health monitoring and ambient assisted living have limitations due to accuracy, affordability, unobtrusiveness, and privacy concerns. Our research aims to address the inconsistent and inaccurate documentation of respiratory and heart rate, the most critical vital signs that can indicate severe pathology, by proposing contactless sensing technique appropriate for patients with respiratory distress in clinical settings. The thesis also presents MechanoBeat, a 3D printed mechanical tag that oscillates at a unique frequency upon user interaction, and an efficient signal processing and deep learning method to locate and identify tags in order to create smart homes and workplaces. Additionally, we introduce MetaScatter, a low-cost, battery-less, 3D-printed micro-structure that can identify everyday objects in complex and dynamic environments. The identification is performed through multi-view convolutional neural networks that fuse dynamic back-scattered signals captured by multiple ultra-wideband radars to identify the meta-reflector tag. Our work highlights the importance of contactless sensing for human-object interaction detection and constructing dynamic and intelligent environments.
Advisor: Tauhidur Rahman