基于CNN+Bi-LSTM和脑电信号的疼痛分类的研究  

Research on Pain Classification Based on CNN+Bi-LSTM and EEG Signal

作　　者：樊楷 李丛煊 刘海燕[1] 

机构地区：[1]北华航天工业学院,河北廊坊065000

出　　处：《大众科技》2020年第11期12-15,22,共5页Popular Science & Technology

基　　金：国家自然科学基金联合重点项目(61401454、71532014)。

摘　　要：脑电图(Electroencephalogram, EEG)是脑神经细胞的电生理活动在大脑皮层或头皮表面的总体反映,包含了大量的生理和病理信息。近年来,与疼痛相关脑电信号的研究是当前脑认知和临床治疗领域的研究热点和难点问题之一。文章使用卷积神经网络(Convolutional Neural Network,CNN)和循环神经网络(Recurrent Neural Network,RNN)结合的CNN+Bi-LSTM网络算法对疼痛和不痛的脑电信号进行二分类,准确率达到了97.1%,与此同时precision、recall、f1-score分别达到了97%、97%、97%。证明了两种网络结合对研究疼痛的脑电信号是可行的。Electroencephalogram(EEG)is the overall reflection of the electrophysiological activity of brain nerve cells on the cerebral cortex or scalp surface,and contains a lot of physiological and pathological information.In recent years,the research of pain-related EEG signals is one of the current hotspots and difficult problems in the field of brain cognition and clinical treatment.In this paper,the CNN+Bi-LSTM network combined with Convolutional Neural Network(CNN)and Recurrent Neural Network(RNN)is used to classify painful and non-painful EEG signals with an accuracy of 97.1%.At the same time,Precision,recall,and f1-score reached 97%,97%,and 97%respectively.It is proved that the combination of the two networks is feasible for studying the EEG signals of pain.

关 键 词：疼痛分类 脑电图 CNN Bi-LSTM 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]