Toileting Expression Detection System for Children with Disabilities Based on Feature Extraction Using Support Vector Machine

Children with disabilities often struggle to express their need to use the toilet, leading to health issues and inappropriate toileting behaviors. This study aims to identify toileting needs based on the facial expressions of children with disabilities using a camera system. Images captured during school activities were analyzed using 51 facial landmark points, focusing on angles, distances, and inclinations. The TOP5 and TOP10 datasets, based on features with the highest Pearson correlation values, were used in a Support Vector Machine (SVM) classification. The TOP5 model achieved a 96% accuracy using parameters C=25 and gamma=0.001, and 5-fold cross-validation. Despite strong performance, challenges such as data collection difficulties and misclassification errors need to be addressed. Increasing dataset diversity by involving subjects with various disabilities is essential for system improvement.