Emotion Recognition: A New Tool for Healthcare Using Deep Learning Algorithms

Emotion recognition is a challenging task, but it has many potential applications in areas such as human–computer interaction, healthcare, and marketing. The capacity to record in-the-moment electrical activity in the brain makes the use of EEG brainwave data as a possible signal for emotion identification extremely promising. This study examined the use of deep learning algorithms for EEG brainwave data-based emotion classification. We obtained the “EEG Brainwave Dataset: Feeling Emotions” dataset from Kaggle to carry out our study. This dataset consists of EEG recordings made across three-minute periods of pleasant, neutral, and negative emotional states from two persons (one male and one female). Using a Muse EEG headband, dry electrodes were placed on the TP9, AF7, AF8, and TP10 EEG locations to record the brainwave waves. Three deep learning algorithms were used throughout our training process: a deep neural network (DNN), a long short-term memory (LSTM) network, and a gated recurrent unit (GRU) network. The DNN achieved the highest accuracy of 98.44%, followed by the LSTM network with 97.5% accuracy and the GRU network with 97.18% accuracy. We also calculated other metrics such as precision, recall, and F1-score. The results showed that all three algorithms were able to achieve high accuracy in classifying emotions from EEG data. The LSTM network had the highest precision for the positive state (1.00), followed by the DNN (0.99), and the GRU (0.98). The GRU network had the highest recall for the negative state (0.96), followed by the DNN (0.97) and the LSTM (0.95). The DNN had the highest F1-score for all three states.