ASD Children Gait Classification Based On Principal Component Analysis and Linear Discriminant Analysis

Abstract

The aim of this study is to explore the potential of the markerless-based gait features using both unsupervised and supervised algorithms namely the Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) as feature extraction techniques. Firstly, a depth camera that created the three-dimensional (3D) skeleton image of the subject upon detection of movements by the motion sensor is used as data acquisition device in acquiring the walking gait of 30 TD and 23 ASD group. Next, the extracted gait features are translated into two categories named as Direct Joint Feature (DIR) and Reference Joint Feature (REF) features. To evaluate the effectiveness of PCA and LDA as feature extraction, Support Vector Machine (SVM), Naïve Bayes Classifier (NBC) and Artificial Neural Network (ANN) are chosen as classifiers. Overall results showed that LDA is the most suitable feature extraction with accuracy of 99.33% using NBC as classifier with the DIR feature dataset as inputs.