A support vector machine algorithm to extract gait phases from accelerometer data
Date of Issue2018
School of Electrical and Electronic Engineering
The accurate detection of gait events is crucial for clinical gait analysis. However, much of the research done so far has been for indoor experimental conditions, which are vastly different from realistic human gait. As such, resulting algorithms gathered from such studies become less useful and reliable. To date, numerous algorithms developed have had much success in accurately detecting heel-strike events. For toe-off detection however, results have not been as accurate. Thus, the purpose of this study is focused on accurate toe-off detection, although limited heel-strike detection has also been attempted. Gait detection is done using a Support Vector Machine (SVM) algorithm using the MAREA dataset as training data. MAREA dataset includes various experimental settings that simulate real world, dynamic human gait. The main findings are: gait detection in indoor conditions are most accurate, and that more work still needs to be done for the SVM to be able to deal with variation of inclination in gait detection. Overall, the SVM classifier developed is simple and can perform in real time with accurate detection for toe-off gait events in comparison with other gait event detection algorithms.
DRNTU::Engineering::Electrical and electronic engineering
Final Year Project (FYP)
Nanyang Technological University