Identifying Stable Patterns over Time for Emotion Recognition from EEG (1601.02197v1)

Abstract: In this paper, we investigate stable patterns of electroencephalogram (EEG) over time for emotion recognition using a machine learning approach. Up to now, various findings of activated patterns associated with different emotions have been reported. However, their stability over time has not been fully investigated yet. In this paper, we focus on identifying EEG stability in emotion recognition. To validate the efficiency of the machine learning algorithms used in this study, we systematically evaluate the performance of various popular feature extraction, feature selection, feature smoothing and pattern classification methods with the DEAP dataset and a newly developed dataset for this study. The experimental results indicate that stable patterns exhibit consistency across sessions; the lateral temporal areas activate more for positive emotion than negative one in beta and gamma bands; the neural patterns of neutral emotion have higher alpha responses at parietal and occipital sites; and for negative emotion, the neural patterns have significant higher delta responses at parietal and occipital sites and higher gamma responses at prefrontal sites. The performance of our emotion recognition system shows that the neural patterns are relatively stable within and between sessions.

• The paper demonstrates that DE features with GELM achieve 91.07% accuracy on EEG-based emotion recognition using DEAP and SEED datasets.
• It employs advanced feature extraction, smoothing, and dimensionality reduction to effectively analyze multi-frequency EEG data.
• Findings highlight the potential for adaptive, stable EEG models in affective computing, enabling more reliable emotion-aware interfaces.

Identifying Stable Patterns over Time for Emotion Recognition from EEG

This paper addresses the identification of stable patterns in EEG signals for emotion recognition, an area crucial for advancements in affective computing and emotion-attentive interfaces. The authors employ a machine learning framework to evaluate EEG stability over time, focusing particularly on discerning specific neural patterns associated with various emotional states.

Methodology

The paper leverages both the DEAP dataset and a newly developed dataset called SEED, which offers the novelty of session-based data to facilitate stability evaluations over time. The research explores several feature extraction and classification techniques, including DE, DASM, RASM, ASM, and DCAU features, with classifiers like SVM and Graph regularized Extreme Learning Machine (GELM) applied to evaluate performance.

Each EEG feature is scrutinized across frequency bands, including delta, theta, alpha, beta, and gamma. Feature smoothing methods, such as the linear dynamic system (LDS) approach, are evaluated against conventional methods. Dimensionality reduction techniques like PCA and MRMR are employed to manage feature space complexity.

Results

The analysis demonstrates that DE features consistently yield superior classification accuracies, notably outperforming traditional PSD features. The GELM classifier, specifically under the DE features, achieves a noteworthy average accuracy of 91.07% on the SEED dataset, confirming its efficacy in EEG-based emotion recognition.

The paper also uncovers stable EEG patterns over time, showcasing neural activation variances across sessions. The lateral temporal regions, particularly active in beta and gamma bands, correlate with positive emotions, whereas negative emotions exhibit heightened gamma responses at prefrontal sites and delta responses at parietal regions.

Implications and Future Directions

These findings suggest significant implications for real-world affective computing systems, emphasizing the potential for designing adaptive interfaces responsive to users' emotional states. The observed stability of certain EEG patterns over time indicates that durable emotion recognition models can be developed.

However, the performance across sessions implies the necessity for adaptive models to address individual and temporal variations. Future research may explore transfer learning strategies to enhance the model's applicability across broader user demographics and extended timeframes. Additionally, examining the gender, age, and cultural influences on EEG patterns could further refine emotion classification systems.

This comprehensive investigation into EEG stability lays the foundation for more robust and temporally stable models in affective computing, potentially informing advancements in applications ranging from mental health monitoring to interactive entertainment.