UROP Project

Detecting Meal and Exercise Events from Continuous Glucose and Wearable Sensor Data

Type 1 diabetes; continuous glucose monitoring; wearable sensors; machine learning

Research Mentor: Prabin Timilsina,
Department, College, Affiliation: Department of Mechanical & Aerospace Engineering, FAMU-FSU College of Engineering
Contact Email: pt23@fsu.edu
Research Assistant Supervisor (if different from mentor): Dr. Taylor Higgins
Research Assistant Supervisor Email: th22u@fsu.edu
Faculty Collaborators:
Faculty Collaborators Email:

Looking for Research Assistants: Yes
Number of Research Assistants: 1
Relevant Majors: Students from all majors with an interest in programming, data analysis, or digital health are also encouraged to apply.
Project Location: 2525 Pottsdamer Street, Tallahassee, FL 32310-6046 (FAMU-FSU College of Engineering)
Research Assistant Transportation Required: There are bus routes serving the College of Engineering (Innovation Route).
Remote or In-person: Partially Remote
Approximate Weekly Hours: 6-8 hours per week, Flexible schedule (Combination of business and outside of business. TBD between student and research mentor.)
Roundtable Times and Zoom Link:

• Day: Friday, September 4
  Start Time: 12:00
  End Time: 3:00
  Zoom Link: https://fsu.zoom.us/j/95303725357

Project Description

People with type 1 diabetes rely on continuous glucose monitors and insulin pumps to help maintain blood glucose within a safe range. However, everyday events such as meals and exercise can cause rapid changes in glucose levels, and these events are not always recorded or announced on time. Automatically recognizing when a meal or exercise event is likely occurring could reduce missed or delayed event announcements and ultimately support more personalized diabetes management.

In this project, we will investigate whether meal and exercise events can be identified using longitudinal diabetes and wearable-sensor data. We will begin with publicly available and/or simulated datasets containing signals such as continuous glucose measurements, insulin delivery, physical activity, heart rate, and recorded behavioral events. The research assistant will help organize and synchronize these data, visualize patterns surrounding known meal and exercise events, and implement baseline machine-learning approaches for event detection.

Our longer-term goal is to develop two complementary modeling approaches: one that learns an individual's typical patterns of when meals or exercise occur, and another that recognizes short-term physiological and behavioral patterns suggesting that an event may be occurring in real time. We will also investigate whether models personalized to an individual perform better than models trained across multiple individuals. This work will provide a foundation for future personalized, context-aware diabetes technologies.

Research Tasks: • Conduct a literature review on meal detection, exercise detection, and behavioral inference using glucose and wearable-sensor data.
• Help identify and document appropriate publicly available diabetes time-series datasets.
• Organize, clean, and synchronize time-series signals collected at different sampling frequencies.
• Implement and evaluate simple baseline classification models.
• Maintain clear documentation of data-processing and analysis procedures.
• Present findings through a research poster, presentation, and potentially contribute to a scientific manuscript.

Skills that research assistant(s) may need: Recommended: Basic programming experience in Python, MATLAB, or a similar language; familiarity with data analysis, statistics, or machine learning; experience working with spreadsheets or tabular data.

Required: Strong organization skills, willingness to learn independently, attention to detail, and reliable communication. The student should be comfortable troubleshooting problems and documenting their work.

Prior experience with diabetes, biomedical data, wearable sensors, or advanced machine learning is not required.