Electroencephalography Based Fusion Two-Dimensional (2D)-Convolution Neural Networks (CNN) Model for Emotion...

Area of Science:

• Affective computing
• Biomedical signal processing
• Machine learning for emotion recognition

Background:

• Accurate human emotion classification is crucial for human-computer interaction and mental health.
• Existing methods often rely on single-modality data, limiting classification performance.
• Multimodal approaches integrating physiological signals show promise for improved emotion detection.

Purpose of the Study:

• To enhance human emotional classification accuracy using a novel convolution neural network (CNN) model.
• To propose a multimodal data fusion method for emotion classification.
• To investigate the combined use of electroencephalogram (EEG) and galvanic skin response (GSR) signals.

Main Methods:

• Preprocessing of galvanic skin response (GSR) signals using the zero-crossing rate.
• Wavelet transform preprocessing of electroencephalogram (EEG) signals for simultaneous time-frequency analysis.
• Development and hyperparameter tuning of a CNN model for effective EEG feature extraction.
• Fusion of preprocessed EEG and GSR signals for multimodal emotion classification.

Main Results:

• The proposed multimodal approach achieved a classification accuracy of 73.4%.
• Significant performance improvement was observed compared to existing state-of-the-art models.
• The CNN model effectively extracted relevant features from EEG signals.

Conclusions:

• Combining EEG and GSR signals with a tuned CNN model significantly improves emotion classification accuracy.
• The proposed multimodal strategy offers a robust method for emotion analysis.
• This approach has potential applications in affective computing and personalized user experiences.