Deep Learning-based Advanced Emotion Sensing (DL-AES) Model from raw EEG signals

Human-computer interaction (HCI) relies heavily on emotion detection, and assessing human behavior according to emotional states is a crucial study area. By examining physiological or non-physiological information, emotion recognition aims to detect people's emotional states automatically optimization. The global and local features that are derived from the electroencephalogram (EEG) data that is produced after stimulation cannot be fully used by the traditional emotion classification techniques. Thus, this research develops an Advanced Emotion Sensing Model based on Deep Learning using raw EEG data. The Band-pass filtering approach is used as a preprocessing tool on EEG data to find significant brain wave patterns. Determining frequency band power features (delta, theta, alpha, and beta) captures complex inter-band interactions that are crucial for identifying emotional states. Graph Convolutional Neural Networks (GCNN) is used to identify positive, negative, and neutral emotions using the features that have been acquired. The public SEED dataset (SJTU Emotion EEG Dataset), which was acquired by using visual and aural stimuli in sequences from Chinese emotional videos, served as the EEG dataset for the study. The proposed model outperformed the existing models with accuracy of 96%, but at a considerable processing cost because of the algorithm's optimization features. This paper advances the field of emotion sensing in EEG by highlighting the efficacy of various models, the significance of feature selection, and the moral application of these technologies in real-world scenario.