Activity detection and analysis on Android

• Main Author: Ong, Xian Hui
• Other Authors: Lee Bu Sung
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Pattern recognition
• Online Access: http://hdl.handle.net/10356/62060
• Institution: Nanyang Technological University

Tracking and learning the activity patterns of an individual is vital when providing healthcare and awareness to the needy such as the elderly or healthcare patients. In this project, unobtrusive detection of the person activities in outdoor environment is implemented through an Android smartphone device. Common activities such as Running, Standing, Walking, Falling down as well as Climbing up and down are carried out. Features extracted from the raw inertial sensor data are first collected from the mobile device and subsequently used to build classification models using different machine learning algorithms in WEKA Analyzing Tool. Different algorithm and approaches are explored and analysis is carried out to determine which approaches achieve the highest accuracy. The most effective method will then be integrated into the system design. The evaluation results of the experiments show that Decision Tree algorithm achieved the highest accuracy result when conducted on 7 activities performed by users. The supervised method achieved 80.5133% when conducted on Author’s individual data and 73.455% when conducted on multiple users’ collected data using 10- cross validation. An “activity” application was developed on the Android platform and a real-time data transmission system was implemented to conduct analytics and data analysis to backend user. The current location of the user’s will also be recorded in the System.