Human activity recognition (HAR) is a challenging problem that involves classifying human actions based on sensor data. Diffusion models are a promising approach for tackling such challenges by learning underlying patterns in the data. However, traditional diffusion models have limitations in handling discrete data, which is common in HAR datasets. To address this, we developed a tabular diffusion model that can handle discrete data and applied it to the UCI HAR dataset. Our tabular diffusion model achieved higher classification accuracy compared to the vanilla diffusion model and was comparable to current state-of-the-art models. The findings have significant implications for the potential of diffusion models in HAR as well as the development of intelligent systems for monitoring human behavior in real-world scenarios.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Trac, Cindy
• McCarthy, Dillon
• Buasai, Sirut
• Dykstra, Jason

Subject

• Project-Based Learning
• Health
• Computing
• Innovation
• Well-Being
• Industry

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-042523-162420
• 105021

Keyword

• machine learning
• human activity recognition
• tabular diffusion
• gramian angular fields
• generative models
• diffusion models

Advisor

• Rundensteiner, Elke A.

Year

• 2023

UN Sustainable Development Goals

• 3 - Good Health and Well-being
• 9 - Industry, Innovation and Infrastructure

Date created

• 2023-04-25

Resource type

• Major Qualifying Project

Major

• Computer Science

Source

• E-project-042523-162420

Rights statement

• In Copyright

Last modified

• 2023-06-22