Emotional State Prediction through Handwriting Analysis

Abstract

Negative Emotions can cause serious health issues and detection of them through daily activities is good for wellbeing .Handwriting and drawings are the activities which requires no specific devices only pen and paper is enough for these activities. Also some devices such as tablets which makes human-machine interface more easily is also good to collect useful information from handwriting and drawing activities. This research investigates the potential of handwriting as an early diagnostic tool for 3 basic emotions Depression, Stress and Anxiety. We study the effectiveness of offline attributes as well as online of handwriting in characterizing the presence or absence of these negative emotions. These online features considered in our study are aimed to extract the pen pressure, in air and on paper movements and compute those measurements which are related to timing and offline features automatically by applying CNN. A key advantage of 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 Depressed, stressed and anxious subjects and the realized results demonstrate the effectiveness of offline features and online features in predicting toxic emotions from handwriting and drawings