EE/CSE Multimodal Wearable AI

Machine learning has enabled wearable devices such as gloves, sleeves, and electronic skins to capture body movements and support applications including gesture recognition, activity tracking, and object interaction understanding. However, many existing wearable systems remain bulky, rigid, or difficult to integrate seamlessly into everyday garments. They also often rely on large amounts of labeled, user-specific training data, limiting their scalability across different users, body types, and wearing conditions.

In contrast, printed and textile-integrated sensors, including resistive, capacitive, and stretchable sensing structures, offer a lightweight, soft, and highly conformal platform for smart textiles. These sensors can be fabricated directly on flexible substrates or integrated into fabrics, enabling customizable sensing layouts that naturally follow the contours and deformation of the human body. This project aims to combine smart textile sensing with an unsupervised meta-learning framework to enable user-independent and data-efficient recognition of hand and body movements.

By leveraging wearable AI, the system will learn transferable representations from multimodal textile sensor signals and adapt to new users and movement patterns with minimal labeled data. Instead of requiring dense sensor arrays or complex hardware, a small number of textile-integrated sensors can capture coordinated motion across multiple joints, supporting soft, comfortable, and low-power wearable intelligence for everyday interaction and health-related applications.