Human Activity Classification Using Smartphone Sensors

Presentation author(s)

Ha H Truong ’22, Ho Chi Minh City, Vietnam

Major: Computer Science


As smartphones become a must-have item in our pockets, they also become an invaluable information source about our daily life. Being able to recognize human activities through smartphones’ passive motions, for instance, is a solid foundation to analyze individuals’ physical health, detect accidents, and even predict possible health risks. That task, however, is quite challenging due to the variations of motions belonging to the same action (resulting from the smartphone’s position, the user’s personal habit, the surrounding environment, and so on). My project aims to preprocess those noised motions from different users, find a pattern between them, and classify each to categories of common activities. By applying Decision Tree, Support Vector Machine, and Neural Network, I built 3 data mining models, compared the results, and analyzed factors causing the differences. All of my models are trained and tested on over 10,000 instances (each with a 561-feature vector and 4 more attributes) from “Human Activity Recognition Using Smartphones” - an experiment conducted in 2015 with 30 volunteers (aged 19 to 48).


Eyad Said

This site uses cookies to improve your experience. Read our Web Privacy Policy for more information.

Got it! ×