Evaluating Cultural Impact on Subject-Independent EEG-Based Emotion Recognition Across French, German, and Chinese Datasets

Anshul Sheoran, Camilo Valderrama

Confirmation: I have read and agree with the IEEE BHI 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: Emotion Recognition, Deep Neural Networks, Attention Mechanism, Cultural influence

Abstract: Culture influences emotional expression and recognition, affecting how individuals perceive and regulate emotions. Given this effect of cultural background, previous studies have suggested that incorporating demographic information can enhance emotion recognition in Electroencephalography (EEG) based approaches. However, until now, most studies have focused on improving prediction accuracy, ignoring the extent to which cultural factors impact EEG-based emotion recognition. To address that gap, this study investigates how cultural factors impact emotion prediction by using a stacking model that combines attention mechanism layers with multinomial logistic regression. The attention mechanism layer focused on detecting the cortical areas in which the model paid more attention to predicting the emotions, while the logistic regression analyzed how the cultural factors affect the odds of accurately predicting emotions. To test our model, we used EEG data capturing three emotions (negative, neutral, and positive) from 31 subjects of three nationalities: 15 Chinese, 8 French, and 8 German. Our approach achieved accuracies of 77.3%, 73%, and 65% for recognizing the emotions in the Chinese, French, and German subjects, respectively. Our approach revealed that incorporating cultural information increases the odds of predicting positive emotions for Chinese subjects and negative emotions for French and German subjects. Moreover, French and German subjects exhibited similar neural patterns across emotions, indicating a closer cultural alignment between these groups. Our findings highlight the critical role of cultural context in emotion recognition models. This inclusion not only improves emotion prediction accuracy for subject-independent approaches but also promotes inclusivity and ethical practices in emotion recognition systems.

Track: 1. Biomedical Sensor Informatics

Registration Id: MPNNZ2Y7XX4

Submission Number: 43