Handwriting-based emotion recognition using GNN

Abstract

Emotion detection is essential to know human behavior, but handwriting,a quite unique personal form of expression, remains underexplored in this context. This paper introduces a singular approach using Graph Neural Networks (GNNs) to locate feelings like depression, anxiety, and stress from handwriting. Utilizing the EMOTHAW dataset, we focus on dynamic capabilities such as ‘Azimuth‘, ‘Altitude‘, and ‘Pressure‘, which display significant versions and offer higher insights as compared to the historically used stroke range. After preprocessing the data, this was formatted in such a way that could be used for the model and turned into a version that was implemented to seize complex spatial and temporal relationships in handwriting patterns. The proposed method demonstrates the capability of handwriting as a sturdy modality for emotion detection and gives valuable insights into emotion-centered projects.