Aakriti Adhikari

Abstract: Personalized healthcare monitoring at home is essential for applications like fall detection, post-surgery recovery, and the early diagnosis of health conditions. While wearable sensors are common, they are often cumbersome, expensive, and unreliable due to compliance issues, particularly among elderly users. Vision-based systems address some of these challenges but come with privacy concerns and performance limitations in low-light conditions or when occlusion occurs. Millimeter-wave technology, integrated into ubiquitous devices like 5G home wireless routers, offers a promising solution for contactless and reliable health monitoring, overcoming the limitations of current systems. In this talk, I will share my dissertation work on developing deep learning models for millimeter-wave sensing to enable healthcare applications. I will highlight MiShape, which employs conditional Generative Adversarial Networks (cGANs) to estimate postures from millimeter-wave signals by generating high-resolution human silhouettes and predicting 3D joint positions. I will also showcase my work in contactless sleep monitoring and discuss its potential use for in-bed patient monitoring, paving the way for advanced, contactless healthcare solutions.