A Noise-Resilient and Energy-Efficient AI Model for Eeg-Based Emotion Recognition Induced by Multimedia Stimuli

The rise in mental health issues such as anxiety and depression in today's society highlights the need for new and noninvasive ways to monitor emotions in real time to help with early intervention efforts. EEG signals have emerged as a useful tool due to their ability to understand human emotion by tracking brain activity instantly. However, current methods face two challenges: noise handling and energy efficiency, which limits their usefulness in real-world scenarios. Traditional machine learning methods such as Support Vector Machines (SVM) face difficulties in capturing temporal relationships; whereas CNN models are computationally demanding and not ideal for realtime emotion recognition tasks despite their adept feature extraction capabilities. To overcome these obstacles, we have created a model that combines Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) layers to improve temporal feature extraction while prioritizing the model's energy efficiency and its robustness against noise. Furthermore, our model is trained on EEG data elicited by diverse multimedia stimuli to develop an emotion detection system that is suitable for real-life scenarios and achieves an overall accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 0. 6 \%}$</tex> across eight emotional categories.