Development and Performance Analysis of Gait Classification Algorithm : Application to Orthotic Bracing
No Thumbnail Available
Date
2016-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Orthotic bracing are fitted to improve the quality of a deviating gait cycle, nevertheless recent
studies show that gait cycle deviations are seen due to bracing [4, 36]. In order to correct the
deviation, adjustments should be made in all stages of the process to which it requires evaluating
the deviation. So the objective of this paper is to develop and analyze the performance of an
algorithm for gait classification, to assess the deviation of gait cycle due to Orthotic bracing.
Acceleration data was taken as an input from Inertial Motion Sensor attached at the foot of a
healthy subject. The algorithm was developed using Model based approach, by considering four
machine learning classifiers. The performance of the algorithm was analyzed using the
performance indicators.
The performance analysis illustrates that using categorical class type gives a relatively higher
accuracy with a maximum accuracy of 86%. The performance related to the classifier on the
other hand illustrates that Tree Bagger Classifier and Support Vector Machine classifier provide
a higher accuracy with a balanced true rate. The analysis further indicate that the gait cycle of the
braced side of an Ankle bracing is relatively expressed using the extracted features, it was
expressed on average 58%. While the effect of the bracing on the unbraced side of the leg are
shown more with Knee bracing.
Keywords: Gait classification, Feature extraction, Inertial motion sensor, Machine learning,
Orthotic bracing, Performance analysis, Confusion matrix, hold out cross valdation.
Description
Keywords
Gait classification, Feature extraction; Inertial motion sensor; Machine learning; Orthotic bracing; Performance analysis; Confusion matrix; hold out cross valdation.