Comparison between Linear Discriminant Analysis and Singular Value Decomposition for PD gait classification

In this study, the effectiveness of Linear Discriminant Analysis (LDA) as feature extraction and dimensionality reduction is evaluated and compared with Singular Value Decomposition (SVD) for gait recognition of Parkinson Disease subjects as compared to normal subjects. Here, three feature vectors of gait namely basic, kinetic and kinematic features are extracted and analysed using LDA and leave one out (LOO) recognition method. Next, ANN classifier is used to compare the performance of LDA versus SVD. Initial findings confirmed that the gait pattern between PD and normal subjects can be classified using the three feature vectors with kinematic features outperformed basic spatial-temporal and kinetic feature vectors with 86.71% with ANN as classifier and 87.5% using LOO recognition technique.