A Data-Driven Adaptive Emotion Recognition Model for College Students Using an Improved Multifeature Deep Neural...

Area of Science:

• Neuroscience and Artificial Intelligence
• Affective Computing and Machine Learning

Background:

• College students face increasing academic, social, and personal pressures, leading to significant emotional fluctuations.
• Traditional emotion assessment methods (surveys, interviews) are subjective, time-consuming, and lack data authenticity.
• Objective physiological data, specifically electroencephalogram (EEG) signals, offer a more reliable measure of emotional states.

Purpose of the Study:

• To develop an accurate and efficient method for recognizing college students' emotions using EEG signals.
• To leverage deep neural networks (DNNs) for classifying EEG data and determining emotional states.
• To enhance emotion recognition by employing feature extraction and fusion techniques.

Main Methods:

• Collected electroencephalogram (EEG) data from college students to capture brain activity related to emotions.
• Extracted various features from the EEG signals to comprehensively represent the data.
• Utilized autosklearn for feature fusion, integrating multiple feature sets.
• Employed a deep neural network (DNN) to classify the fused features and predict emotional states.

Main Results:

• The proposed method demonstrated effectiveness on public datasets for emotion recognition.
• Achieved a high accuracy rate exceeding 88% in classifying emotional states from EEG data.
• The feature fusion technique using autosklearn improved the comprehensiveness of EEG data representation.

Conclusions:

• The developed EEG-based emotion recognition system is feasible for real-world application in college student management.
• This objective approach overcomes the limitations of traditional subjective methods.
• The findings support the use of advanced machine learning techniques for monitoring and supporting student mental well-being.