Publication

Abstract : Feelings are the foundation of the human experience which influences choices, interpersonal interactions, and overall well-being. Emotional intelligence is necessary for mental health, emotional growth, and effective communication. In the modern world, the study of human emotions is becoming more and more popular. Because the signals produced by the brain are unstable, developing electronic models to identify emotional states from EEG data is challenging. In this study, we propose a deep learning framework-based efficient technique for EEG data analysis developed and collected from the DEAP dataset. We also use feature selection approaches to identify characteristics relevant to the cognitive search task, based on discoveries from neuroscientific research. The multimodal DEAP dataset consists of audiovisual recordings, physiological signals, and self-reported emotional assessments obtained from exposure to a range of multimedia stimuli includes a huge quantity of data. The data set was written by the scientific community and includes the EEG recordings of 32 participants. Each subject's data was obtained from 32 EEG sensors, while the remaining 8 sensors were obtained from pupils and electromyography. The suggested study takes into account “Support Vector Machines (SVMs), Random Forests, Convolutional Neural Networks (CNNs), and Long Short-Term Memory (LSTM) networks” in addition to a range of supervised and unsupervised learning techniques. The advantages and disadvantages of each approach are explored, with a focus on how crucial it is to select the best approach depending on the type of EEG data and the particular emotional awareness task at hand. Our established model effectively categorized emotions into two main groups: arousal (the strength of the emotion) and valence (the pleasantness of the emotion). Surprisingly, our model was able to identify between these high and low valence, as well as high and low arousal states, with an average accuracy of over 92%. In addition, our model was able to classify emotions into four different groups: peak valence—slight arousal (PVSA), peak valence—peak arousal (PVPA), minor valence—slight arousal (SVSA), and slight valence—peak arousal (SVPA). The accuracy in this more thorough categorization was still very good at 84%. This degree of precision demonstrates the model's ability to identify and discriminate between complex emotional states, highlighting its potential in a range of emotion detection applications.