基于脑电信号的聋人情绪识别研究  

Research on the deaf emotion recognition based on the electroencephalogram signals

作　　者：毛泽民 康巧菊 宋雨[2] 袁龙灿 MAO Zemin;KANG Qiaoju;SONG Yu;YUAN Longcan(Technical College for the Deaf,Tianjin University of Technology,Tianjin 300384,China;Tianjin Key Laboratory for Control Theory and Applications in Complicated Systems,School of Electrical Engineering and Automation,Tianjin University of Technology,Tianjin 300384,China)

机构地区：[1]天津理工大学聋人工学院,天津300384 [2]天津理工大学电气工程与自动化学院天津市复杂系统控制理论及应用重点实验室,天津300384

出　　处：《天津理工大学学报》2024年第2期92-99,共8页Journal of Tianjin University of Technology

基　　金：国家自然科学基金(62103299);天津理工大学教学基金(ZX21-08)。

摘　　要：针对聋人大学生容易出现情感表达障碍和情绪识别困难等问题,文中提出了基于脑电(electroencephalogram,EEG)信号的聋人情绪识别方法。首先对采集到的EEG信号用5阶巴特沃斯滤波器将其分割为5个频带,提取各频带的功率谱密度(power spectral density,PSD)和微分熵(differential entropy,DE)特征,利用支持向量机(support vector machine,SVM)进行分类实验。分析了聋人在不同情绪状态下的脑地形图,得出与情绪处理相关的关键脑区。最后根据关键脑区分布,提出了聋人情绪识别的关键通道选择模式并进行了分类实验。实验结果表明,与聋人情绪相关的主要频带为Beta和Gamma频带,且DE特征在聋人情绪识别中具有较大优势,在全频带62通道的平均识别准确率可达到97.57%;与聋人情绪处理相关的主要脑区为前额叶和枕叶;DE特征在10通道的识别准确率达到96.14%,可与62通道分类结果相媲美。研究结果可为开发应用于聋人情绪识别的可穿戴EEG设备提供参考。As for the deaf college students prone to emotional expression disorders and emotional recognition difficulties,this paper proposes a deaf emotional recognition method based on electroencephalogram(EEG)signals.Firstly,the collected EEG signal was divided into 5 frequency bands by the 5th-order Butterworth filter,the power spectral density(PSD)and differential entropy(DE)features of each frequency band were extracted,and the support vector machine(SVM)was used for classification experiments.Then the brain topography of deaf people in different emotional states was analyzed,and the key brain regions related to emotional processing were derived.Finally,according to the distribution of key brain regions,key channel selection models for emotion recognition of deaf people were proposed and classification experiments were carried out.The experimental results show that the main frequency bands related to deaf emotional processing are Beta and Gamma frequency bands,and the DE feature has a great advantage in deaf emotion recognition,and the average recognition accuracy of 62 channels in the whole frequency band can reach 97.57%;The main brain regions related to emotional processing of the deaf are the prefrontal and occipital lobes;The recognition accuracy of the DE in 10-channel reaches 96.14%,which is comparable to the 62-channel classification results.The research results can provide a reference for the development of wearable EEG devices for deaf emotion recognition.

关 键 词：情绪识别 聋人 脑电 微分熵 通道选择 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]