Abstract:
This research presents a novel insight on gait disorder detection using transfer learning
algorithms on sensor-acquired data based on the implementation of popular Convolutional
Neural Network (CNN) models. So, the research basically deals with Deep Learning. The
dissertation proposes the use of pressure sensors to extract heatmap images during gait, which
are then trained and tested in various classification algorithms for gait abnormality diagnosis
and detection. Gait is a biological and scientific study of body movement and locomotion that
emphatically serves as a reliable parameter for inspecting the human body’s neuromuscular
and skeletal systems. To build a convenient and precise classification system for possible
application, synthetic data was generated in multiple preexisting CNN models, which were
then evaluated using conventional performance metrics. The proposed notion yielded
experimental findings that showed higher accuracies for all transfer learning schemes tested,
with the Vgg16 model achieving a notable accuracy of 97.15%. As a result, the analysis
demonstrated not only a significant performance in terms of accuracy, but also reduced
complexity and computing time, making the approach efficient yet effective. A detailed
comparative analysis of performance with all other algorithms was carried out in terms of
accuracy, precision, recall, and F-1 score
Description:
Supervised by
Mr. Mirza Muntasir Nishat,
Assistant Professor,
Co-Supervisor,
Mr. Fahim Faisal,
Assistant Professor,
Department of Electrical and Electronic Engineering (EEE),
Islamic University of Technology (IUT),
Board Bazar, Gazipur-1704, Bangladesh.
This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Electrical and Electronic Engineering, 2022.