Emotion Detection Using Deep Normalized Attention-Based Neural Network and Modified-Random Forest.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Machine Learning

Background:

• Emotion detection is crucial for applications like biometric security and human-computer interaction (HCI).
• Electroencephalogram (EEG) signals offer high sensitivity to emotional state changes, making them valuable for emotion detection.
• Manual analysis of EEG signals is time-consuming, and existing AI algorithms lack sufficient accuracy.

Purpose of the Study:

• To develop an accurate and efficient method for emotion detection from EEG signals.
• To address the limitations of current data mining algorithms in EEG-based emotion recognition.
• To propose a novel deep learning architecture for feature extraction and classification.

Main Methods:

• Proposed a Deep Normalized Attention-based Residual Convolutional Neural Network (DNA-RCNN) for discriminative feature extraction from EEG signals.
• Incorporated attention modules within the DNA-RCNN to enhance feature representation and performance.
• Utilized a modified-random forest (M-RF) classifier with an empirical loss function for precise emotion classification.

Main Results:

• The proposed DNA-RCNN effectively extracts relevant features from EEG signals.
• Attention modules in the network contribute to consistent and improved performance.
• The M-RF classifier, combined with the proposed network, achieves precise classification by minimizing prediction errors.

Conclusions:

• The developed system demonstrates superior performance in emotion detection from EEG signals compared to existing methods.
• The combination of DNA-RCNN and M-RF offers an effective solution for accurate and efficient EEG-based emotion recognition.
• The study confirms the effectiveness of the proposed approach for real-world emotion detection applications.