摘要
利用功能性近红外光谱(f NIRs)技术实现了对不同情绪状态的识别。采集了15名受试者在6种情绪种类图片刺激下的f NIRs信号以及唤醒度、愉悦度评价数据。为了实现对受试者情绪状态的分类评估,采用支持向量机(SVM)和基于支持向量机的递归特征筛选(SVM-RFE)算法来筛选参数并设计情绪状态的分类器。结果表明在多种情绪种类图片刺激下,受试者出现了显著的功能响应曲线,并且在唤醒度、愉悦度和情绪种类三个分类目标上分别实现了81%、78.78%和68%的平均分类正确率。同时发现唤醒度和愉悦度的敏感特征主要出现在眶额叶皮层和背外侧皮层,且近似熵是反映情绪状态变化的有效指标。因此采用f NIRs能够基本实现对人体情绪状态的识别。
In order to investigate the human emotional state recognition, the functional near-infrared spectroscopy （fNIRs） technique is applied to measure hemodynamic signals of 15 participants who are requested to see six types of pictures, and the participants have to complete 7-point rating scale of valence and arousal after every picture stimulus. The support vector machine （SVM） and support vector machine based recursive feature elimination （SVMRFE） algorithm are applied to design classifiers. Under different emotional image stimulus, the hemodynamic signals of some participants show significant neural response. With the target classification based on valence, arousal and emotion category, the accuracy is 81%, 78.78% and 68%, respectively. The 5th and 6th channels for fNIRs measurement are significantly sensitive to arousal and valence state, and the two channels are located at orbitonfrontal cortex and dorsolateral prefrontal cortex regions. Besides, it is found that the entropy of fNIRs can reflect the variation in emotional state effectively. The results suggest that fNIRs can be used for recognition of human emotional state.
出处
《光学学报》
EI
CAS
CSCD
北大核心
2016年第3期156-166,共11页
Acta Optica Sinica
基金
国家自然科学基金青年基金(71201148)
国家973计划(2011CB711000)
中国航天员科研训练中心人因工程重点实验室实验技术课题(9140C770208150C77320
2012SY54B1701)
关键词
医用光学
情绪状态识别
功能性近红外光谱技术
支持向量机
模式识别
medical optics
emotional state recognition
functional near-infrared spectroscopy
support vector machine
pattern recognition
