Wu Wanmeng, Zuo Enling, Zhang Weiya, Meng Xiangjie
School of International Education and Exchange, Changchun Sci-Tech University, Changchun, China.
School of Education, Changchun Normal University, Changchun, China.
Front Public Health. 2024 Nov 26;12:1492375. doi: 10.3389/fpubh.2024.1492375. eCollection 2024.
An increasing prevalence of psychological stress and emotional issues among higher education teachers necessitates innovative approaches to promote their wellbeing. Emotion recognition technology, integrated into educational human-computer interaction (HCI) systems, offers a promising solution. This study aimed to develop a robust emotion recognition system to enhance teacher-student interactions within educational HCI settings.
A multi-physiological signal-based emotion recognition system was developed using wearable devices to capture electrocardiography (ECG), electromyography (EMG), electrodermal activity, and respiratory signals. Feature extraction was performed using time-domain and time-frequency domain analysis methods, followed by feature selection to eliminate redundant features. A convolutional neural network (CNN) with attention mechanisms was employed as the decision-making model.
The proposed system demonstrated superior accuracy in recognizing emotional states than existing methods. The attention mechanisms provided interpretability by highlighting the most informative physiological features for emotion classification.
The developed system offers significant advancements in emotion recognition for educational HCI, enabling more accurate and standardized assessments of teacher emotional states. Real-time integration of this technology into educational environments can enhance teacher-student interactions and contribute to improved learning outcomes. Future research can explore the generalizability of this system to diverse populations and educational settings.
高等教育教师中，心理压力和情绪问题的患病率日益上升，因此需要创新方法来促进他们的幸福感。融入教育人机交互（HCI）系统的情绪识别技术提供了一个很有前景的解决方案。本研究旨在开发一个强大的情绪识别系统，以加强教育HCI环境中的师生互动。
利用可穿戴设备开发了一个基于多生理信号的情绪识别系统，用于采集心电图（ECG）、肌电图（EMG）、皮肤电活动和呼吸信号。使用时域和时频域分析方法进行特征提取，然后进行特征选择以消除冗余特征。采用具有注意力机制的卷积神经网络（CNN）作为决策模型。
与现有方法相比，所提出的系统在识别情绪状态方面表现出更高的准确率。注意力机制通过突出用于情绪分类的最具信息性的生理特征来提供可解释性。
所开发的系统在教育HCI的情绪识别方面取得了重大进展，能够对教师的情绪状态进行更准确和标准化的评估。将该技术实时集成到教育环境中可以加强师生互动，并有助于提高学习成果。未来的研究可以探索该系统对不同人群和教育环境的通用性。
