Abstract:
Parkinson disease (PD) is a degenerative disorder of the central nervous system that influences the movement of muscles causing tremors, slowed movements, impaired balance and rigidity. While sophisticated medical procedures exist to diagnose PD, these methods indicate PD only after the disease has progressed considerably. It has been established in a number of studies that analysis of handwriting and speech can serve as an effective early warning for Parkinson. This research investigates the potential of handwriting as an early diagnostic tool for PD. While traditional techniques to predict PD from handwriting mostly rely on online features like pen pressure, in-air and on-surface movements, we study the effectiveness of offline attributes of handwriting in characterizing the presence or absence of PD. These offline features considered in our study are aimed to extract the textural information in handwriting and include local binary patterns, gray-level co-occurrence matrices and fractal dimensions. A key advantage of such offline attributes is that unlike online features, no specialized hardware is required to compute these features. For classification, we employ Support Vector Machine (SVM) and Artificial Neural Network (ANN). Evaluations are carried out on a benchmark dataset including PD patients and control subjects and the realized results demonstrate the effectiveness of offline features in predicting PD from handwriting.
