Treffer: Optimized node-level capsule graph neural network for subject-independent emotion recognition from EEG signals.

Subject-independent emotion detection using EEG (Electroencephalography) using Vibrational Mode Decomposition and deep learning is made possible by the scarcity of labelled EEG datasets encompassing a variety of emotions. Labelled EEG data collection over a wide range of emotional states from a broad and varied population is challenging and resource-intensive. As a result, models trained on small or biased datasets may fail to generalize well to unknown individuals or emotional states, resulting in lower accuracy and robustness in real-world applications. A Node-Level Capsule Graph Neural Network (NCGNN) is then used to correctly recognize emotions like calm, happy, sad, and furious based on the features that have been collected. Generally speaking, the NCGNN classifier does not provide optimization techniques for adjusting parameters to ensure precise emotion recognition. Hence, propose to utilize the Piranha Foraging Optimization Algorithm (PFOA) to enhance Node-Level Capsule Graph Neural Network, accurately categorize the emotion level. Then, the proposed NLCGNN-SIER-EEG is excluded in Python and the performance metrics like Recall, Accuracy, Precision, Specificity, F1 score and RoC. In the end, the performance of NLCGNN-SIER-EEG technique provides 19.57%, 24.37% and 34.15% high accuracy, 22.12%, 26.82% and 28.52% higher Precision and 23.26%, 28.17% and 29.43% higher recall while compared with existing like Subject-independent emotion recognition based on EEG data using VMD and deep learning (SIER-EEG-VMD-DL), Emotion recognition system based on two-level ensemble of deep-convolutional neural network models (ERS-TLE-DCNN), and human emotion recognition based on EEG data using principal component analysis and artificial neural networks (EEH-HER-ANN), respectively. PLAIN-LANGUAGE SUMMARY: Subject-independent emotion detection using EEG (Electroencephalography) using Vibrational Mode Decomposition and deep learning is made possible by the scarcity of labelled EEG datasets encompassing a wide range of emotions. Labeled EEG data collection over a wide range of emotional states from a broad and varied population is challenging and resource-intensive. As a result, models trained on small or biased datasets may fail to generalize well to unknown individuals or emotional states, resulting in lower accuracy and robustness in real-world applications. The gathered traits are then entered into an NCGNN to correctly detect emotions like calm, happy, sad, and enraged. [ABSTRACT FROM AUTHOR]