A Physiological Approach to Affect Recognition in Children: Insights from EDA

Abstract Objectives:  This study aimed to develop an automated approach for detecting affective states in typically developing children by analyzing their electrodermal activity (EDA) and addressing the challenges and significance of emotion recognition through physiological signals. Method: The study involved 15 children aged 8–12 years whose multimodal emotional responses were recorded using an EmotiBit device during emotion elicitation via short video sequences. Key characteristics were derived from EDA signals using discrete wavelet transform, and classification was executed using a Multilayer Perceptron (MLP) network. Findings: The proposed method achieved an accuracy of 80% in a three-class emotion classification task, demonstrating its effectiveness in recognizing the affective states. Novelty: The proposed methodology has the potential to facilitate the development of sophisticated and intelligent systems that can accurately recognize emotions in typically developing children by using electrodermal activity. Keywords: Emotion recognition, Electrodermal activity (EDA), discrete wavelet transform, machine learning, children, Multilayer Perceptron (MLP)