AI-Enhanced Personalized Step Detection Algorithms for Older Adults Using Wearable Sensors

Current step-counting algorithms in wearable devices are inaccurate for older adults because they were designed and tested on younger populations. This creates measurement errors that undermine physical activity monitoring for the group that needs it most. This project will analyze accelerometer data from older populations to identify gait characteristics such as cadence, walking speed variability, and movement patterns unique to seniors. Machine learning models will be trained and benchmarked against existing algorithms, with an emphasis on accuracy at slower walking speeds and during real-world activities. Additionally, the project will create a reproducible, open-source analytical pipeline compatible with standard wearable devices used in Canadian research. The University of Saskatchewan will gain advanced expertise in wearable sensor analytics and AI-based health monitoring. At the same time, the Oxford Data Science Institute will extend the application of its methods to Canadian aging populations and digital health contexts. The resulting algorithms will directly support the Gerofit application and improve the accuracy of physical activity measurement for older Canadians, contributing to chronic disease prevention and healthy aging initiatives.

Faculty Supervisor:

Daniel Fuller

Student:

Partner:

University of Oxford

Discipline:

Computer science

Sector:

Health and Related Sciences & Technology

University:

University of Saskatchewan

Program: