Enhancing Emotion Recognition Performance Using EEG Signals with Deep Learning Algorithms and Hyperparameter Optimization

Emotions play a crucial role in understanding human nature and enable us to experience and feel what we go through. Humans express their behavioral characteristics through emotions. In addition to or as an alternative to facial image-based emotion recognition, the EEG (Electroencephalogram) method is also employed in emotion recognition tasks. In this study, a novel deep learning approach has been developed to enhance emotion recognition performance using EEG signals. We applied hyperparameter optimization on SEED-IV, DEAP, and DREAMER datasets using our proposed Special Convolutional Model (SCM) and two-dimensional LSTM models. The SCM model achieved an accuracy rate of 64.35% on the SEED-IV dataset, outperforming similar studies in the literature. The two-dimensional LSTM model, on the other hand, achieved a notable accuracy of 58.8%. The success of the SCM model is attributed to hyperparameter optimization performed using the RMSprop optimizer. In tests conducted on the DREAMER dataset, the accuracy rate, which was 28% in the WEKA environment, was increased to 44% through optimization. This study demonstrates the effectiveness of optimizations in deep learning models for enhancing EEG-based emotion recognition performance. The developed approach allows for a deeper understanding of EEG signals and surpasses existing methods in the literature.