Feature Relevance Analysis for Handwriting Based Identification of Parkinson ’s Disease

Abstract

Parkinson’s disease is a severe neurodegenerative disorder that impairs the motor system over time, causing the slowness of speech and movements, as well as abnormal writing abilities due to tremors. Parkinson’s patients are not suitable for all types of PD diagnosis tests due to their physical problems. As a result, a handwriting test can be used to construct an automated diagnostic tool as a potential marker. While traditional techniques focused on the effectiveness of online and offline or combining both features of handwriting from established templates characterizing the presence and absence of PD. In this study, we use the PaHaW dataset to carry out a comprehensive study to assess the optimal set of features that are more informative as a function of the templates from which they are extracted. For this purpose, We extract online and offline features subjects,combined extracted features and employed a feature selection mechanism , such as a genetic algorithm and correlation, to find the most relevant features that describe the presence and absence of PD by achieving an overall accuracy of 77.46% .