Abstract
In recent years, emotion recognition has received increasing attention as it plays an essential role in human-computer interaction systems. This paper proposes a four-class multimodal approach for emotion
recognition based on peripheral physiological signals that uniquely combines a Continuous Wavelet Transform (CWT) for feature extraction, an overlapping sliding window approach to generate more data samples and a Convolutional Neural Network (CNN) model for classification. The proposed model processes multiple signal types such as GalvanicSkin Response (GSR), respiration patterns, and blood volume pressure. Achieved results indicate an accuracy of 84.2%, which outperforms state-
of-the-art models on four-class classification despite of being only based on peripheral signals