基于CNN和LSTM的脑电信号情感识别
EEG based emotion recognition using CNN and LSTM
  
DOI:
中文关键词:  情感识别;卷积神经网络;长短时记忆网络;脑电信号
英文关键词:emotion recognition;convolutional neural network(CNN);long short term memory (LSTM);electroencephalogram(EEG)
基金项目:江苏省重点研发计划(BE2016775)和江苏省研究生科研创新计划(KYCX19_0899,KYCX19_0954)资助项目
作者单位
卢官明 南京邮电大学 通信与信息工程学院,江苏南京210003 
丛文康 南京邮电大学 通信与信息工程学院,江苏南京210003 
魏金生 南京邮电大学 通信与信息工程学院,江苏南京210003 
闫静杰 南京邮电大学 通信与信息工程学院,江苏南京210003 
摘要点击次数: 299
全文下载次数: 105
中文摘要:
      为了提高脑电信号情感识别的准确率,提出了一种基于卷积神经网络(CNN)和长短时记忆(Long Short Term Memory,LSTM)网络的脑电信号情感识别方法。首先,对62个通道的脑电信号进行预处理,并对预处理后的每个通道的脑电信号分别采用一维卷积神经网络提取情感特征。然后,利用LSTM网络在序列上的建模能力,将62个通道的情感特征组成特征序列依次输入到LSTM网络,提取多通道融合情感特征。最后,将LSTM网络输出的多通道融合情感特征输入到全连接层和Softmax分类器,将情感分成积极、中性、消极3种类别。在脑电情感数据集SEED上进行了情感识别实验,取得了88.15%的平均分类准确率。实验结果表明,文中提出的脑电信号情感识别方法的性能优于基于传统人工设计特征及支持向量机(SVM)或深度置信网络(DBN)的其他方法,验证了文中提出方法的可行性和有效性。
英文摘要:
      To improve the classification accuracy of emotion recognition from EEG signals,an EEG based emotion recognition method is proposed by using convolutional neural network (CNN) and long short term memory (LSTM) network.Firstly,the EEG signals from 62 channels are pre processed and the emotion features are separately extracted from the pre processed EEG signals for each channel by using one dimensional convolutional neural network.Then,the emotion features from 62 channels are in turn input into the LSTM network to extract the multi channel fusion emotion features by the modeling ability of LSTM network in sequence.Finally,the multi channel fusion emotion features from LSTM network are input into the full connection layer and Softmax classifier,and the emotion is classified into three categories:positive,neutral and negative.The experiment for emotion recognition is carried out on the SJTU EEG emotion dataset (SEED),and the average classification accuracy is 88.15%.Experimental results show that the performance of the proposed method is better than that of other methods based on traditional hand craft features and support vector machine (SVM) or deep belief network (DBN),thus demonstrating the feasibility and the effectiveness of the proposed method.
查看全文  查看/发表评论  下载PDF阅读器

你是第2142690访问者
版权所有《南京邮电大学学报(自然科学版)》编辑部
Tel:86-25-85866913 E-mail:xb@njupt.edu.cn
技术支持:本系统由北京勤云科技发展有限公司设计

欢迎访问《南京邮电大学学报(自然科学版)》编辑部!