Positive and Negative Emotions Recognition using Multichannel EEG Analysis with Spectral Entropy and Machine Learning approaches

Abstract Emotion recognition is a critical component of human-computer interaction, facilitating more natural and effective communication between users and machines. This study explores the recognition of positive and negative emotions using a multichannel Electroencephalogram (EEG) analysis approach, incorporating spectral entropy and machine learning techniques. Spectral entropy, which quantifies the complexity and irregularity of EEG signals, is employed to extract relevant features from the raw EEG data. These features are then fed into various machine learning models to classify emotions accurately. The research leverages a comprehensive dataset of EEG recordings from participants exposed to emotion-inducing stimuli, ensuring a robust analysis of both positive and negative emotional states. Preprocessing steps, including noise reduction and artifact removal, are meticulously applied to ensure the quality and reliability of the EEG signals. Spectral entropy features are extracted from multiple frequency bands, capturing the nuanced variations in brain activity associated with different emotional states. Several machine learning algorithms, including Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbor (kNN), Decision Tree (DT), and Neural Network (NN), are evaluated for their performance in classifying the emotions. The study also examines the impact of feature selection and dimensionality reduction techniques on the accuracy and efficiency of the classifiers. Cross-validation and other statistical methods are employed to validate the models and prevent overfitting. The preliminary results from the integration of spectral entropy features with advanced machine learning approaches demonstrate an accuracy of 73.7% and f1-scores of approximately 0.5974 to 0.7500 when differentiating between negative, positive and neutral emotional states. The findings suggest that multichannel EEG analysis, when combined with sophisticated feature extraction and classification techniques, holds great promise for developing more intuitive and responsive human-computer interfaces. Future research directions include the application of deep learning models to further improve the robustness and scalability of emotion recognition systems.