基于LSTM-Transformer的脑电情感分析  

EEG emotion analysis based on LSTM-Transformer

作　　者：王安琪 于超 陈胤玮 郗群[2] WANG Anqi;YU Chao;CHEN Yinwei;XI Qun(School of Information Engineering,Gansu University of Chinese Medicine,Lanzhou 730000,China;Information Center,the Second Hospital of Lanzhou University,Lanzhou 730001,China)

机构地区：[1]甘肃省中医药大学信息工程学院,甘肃兰州730000 [2]兰州大学第二医院信息中心,甘肃兰州730001

出　　处：《中国医学物理学杂志》2024年第12期1550-1557,共8页Chinese Journal of Medical Physics

基　　金：甘肃省自然科学基金(20CX9JA145)。

摘　　要：针对传统情感识别方法在处理长期依赖关系时的不足,提出一种结合长短期记忆网络(LSTM)与Transformer模块的脑电情感分析模型,称为LTNet。该模型首先对数据进行预处理,然后将提取的特征输入至LTNet。LSTM模块和Transformer模块独立对输入的序列进行建模,分别从中提取出深层的局部特征和全局特征。通过采用加权融合策略来综合这些特征,最终利用Softmax函数对情绪进行四分类。实验结果显示,在DEAP数据集上进行的五折交叉验证中,LTNet的平均识别准确率达到96.56%,相比于传统机器学习算法和其他深度学习方法提高2.74%~21.31%。An electroencephalogram(EEG)emotion analysis model(LTNet)that combines long short-term memory(LSTM)and Transformer modules is proposed for addressing the shortcomings of traditional emotion recognition methods in dealing with long-term dependencies.After data preprocessing,the extracted features are input into LTNet.LSTM module and Transformer module model the input sequence independently,and from which deep local features and global features are extracted and then fused using a weighted fusion strategy.Finally,Softmax function is used to classify emotions into 4 categories.Experimental results show that LTNet has an average recognition accuracy of 96.56%in the 5-fold crossvalidation on the DEAP dataset,which is 2.74%-21.31%higher than traditional machine learning algorithms and other deep learning methods.

关 键 词：脑电 情绪识别 深度学习 LSTM TRANSFORMER 

分 类 号：R318[医药卫生—生物医学工程] TP391.7[医药卫生—基础医学]