Classification of stress and participation using physiological signals of children during serious game-based therapy

Abstract

The study involves classifying stress and participation of children with special needs and typically developing children using the physiological signals collected during serious game-based therapy. Blood Volume Pulse (BVP), Electrodermal Activity (EDA), and Skin Temperature (ST) physiological signals were collected from 25 children with special needs (dyslexia, intellectual disabilities) and typically developing. The 98 features from the physiological signals were extracted and significant physiological features were found using an independent t-test. The most significant features are selected, and Support Vector Machines (SVM), k-Nearest Neighbors (kNN), Random Forest (RF), Decision Tree (DT), Logistic Regression (LR), Naïve Bayes (NB), and Artificial Neural Networks (ANN) machine-learning methods are used to classify stress/no-stress and participation/no-participation. The highest classification accuracy and F1-Score were obtained as 0.90 and 0.76, respectively, with the RF method using the significant features for stress/no-stress classification. For participation/no-participation, the highest classification accuracy and F1-Score were obtained as 0.80 and 0.70, respectively, with RF method using all features.