Abdullah Ali Alnaqbi

Artificial Intelligence Specialization · Cohort 2025

WiFi Human Activity Recognition

This project explores how we can recognize human activities (like walking, sitting, or falling) using only WiFi signals, specifically Channel State Information (CSI). When a person moves, it affects how WiFi signals behave — and by learning from these changes, we can try to classify the activity taking place. The main challenge in this field is that most available datasets are small and not diverse, which makes models perform poorly on new users or environments or domains. While current approaches use self-supervised learning (SSL), transfer learning, or few-shot learning to adapt across domains, they still struggle when the source data lacks variety. Our project focuses on using diffusion models to generate synthetic CSI data, aiming to increase dataset diversity and improve model generalization.

Mentor: Muhammad Mubashar

Read the final report (PDF)

See it in the program report (PDF)