军事医学2025，Vol.49Issue(11)：823-828,6.DOI:10.7644/j.issn.1674-9960.2025.11.003
基于fNIRSNet的恐惧情绪快速识别技术研究
A technology for fast recognition of fear emotions based on fNIRSNet
李宇睿 1付文亮 2徐东刚 2高波 2邢微微 2左中夫3
作者信息
- 1. 锦州医科大学基础医学院,辽宁 锦州 121001||军事科学院军事医学研究院,北京 100850||辽宁省糖尿病感知功能障碍重点实验室,辽宁 锦州 121001
- 2. 军事科学院军事医学研究院,北京 100850
- 3. 锦州医科大学基础医学院,辽宁 锦州 121001||辽宁省糖尿病感知功能障碍重点实验室,辽宁 锦州 121001
- 折叠
摘要
Abstract
Objective To establish a technology for rapid recognition of fear emotions in order to facilitate the monitoring and interventions of fear emotions among personnel doing unusual jobs based on the fNIRSNet model.Methods Functional near infrared spectroscopy(fNIRS)was used to record the changes of cerebral blood oxygen activities in 50 subjects watching non-thrilling and thrilling videos,whose emotions were assessed based on Self-Assessment Manikin(SAM).Signals of blood oxygen in response to the two types of videos were labeled and preprocessed.Channels with significant differences were identified by calculating the Beta value.The features of data on fNIRS in the activated region of the brain where the channels were located were extracted.The fNIRSNet algorithm was used to establish a fear emotion recognition model,and the accuracy was evaluated using the 50-fold cross-validation method.Results The SAM showed that fear emotions were induced,and the brain region activated by fear emotions was located in the medial prefrontal cortex.The fNIRSNet algorithm could help classify fear emotions with high precision by analyzing the data on the 20 s short time series,with an accuracy of 82.36%.Conclusion This technology for emotion recognition based on the fNIRSNet model is capable of rapid and accurate assessment of fear emotions,which can be used for related monitoring and interventions.关键词
功能性近红外光谱/恐惧情绪/快速识别/情绪识别/深度学习/fNIRSNetKey words
functional near infrared spectroscopy/fear emotion/rapid recognition/emotion recognition/deep learning/fNIRSNet分类
医药卫生引用本文复制引用
李宇睿,付文亮,徐东刚,高波,邢微微,左中夫..基于fNIRSNet的恐惧情绪快速识别技术研究[J].军事医学,2025,49(11):823-828,6.
