EEG-Based Emotion Classification Using Improved Cross-Connected Convolutional Neural Network.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Electroencephalography (EEG) is crucial for advancing human-computer interactions through emotion recognition.
• Developing accurate and efficient emotion classification models remains a significant challenge.

Purpose of the Study:

• To propose an improved deep convolutional neural network (CNN) model for enhanced emotion classification using EEG data.
• To evaluate the model's performance using a non-end-to-end training method combining multi-layer features.

Main Methods:

• A novel deep CNN model was developed, integrating features from bottom, middle, and top convolution layers.
• Four experimental sets with 4500 samples were used to validate the model's performance.
• Feature visualization and scatterplot analysis were employed to assess feature separability.

Main Results:

• The proposed model achieved a high classification accuracy of 93.7%.
• Extracted features demonstrated superior separability compared to other tested models.
• Model performance was not significantly impacted by redundant layers or removal of specific EEG channels.

Conclusions:

• The developed model offers higher accuracy and speed for emotion recognition than existing methods.
• The approach is suitable for applications requiring rapid and precise identification of human emotions.
• This research contributes to the advancement of intelligent human-computer interaction systems.