基于深度学习的运动想象脑电信号识别方法  

Deep learning-based method for recognition of motion imagery EEG signal*

作　　者：宋春宁[1] 盛勇 宁正高 SONG Chunning;SHENG Yong;NING Zhenggao(School of Electrical Engineering,Guangxi University,Nanning 530004,China)

机构地区：[1]广西大学电气工程学院,广西南宁530004

出　　处：《传感器与微系统》2022年第4期125-128,133,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(51767005);广西自然科学基金资助项目(2016GXNSF AA380328)。

摘　　要：在脑电(EEG)信号分析方法中,时频分析方法综合考虑了信号的时间与频率两者的分辨率,同时改善了单纯时间域或频率域分析方法的短板。本实验使用S变换代替短时傅里叶变换将左右手运动想象脑电信号转换为二维时频图像形式,然后构建卷积神经网络-极限学习机(CNN-ELM)模型进行分类。在面对小样本训练数据时模型能力受到限制,提出一种数据增强方法,通过ACGAN对时频图像进行生成,有效丰富了训练样本数量。实验结果表明:CNN-ELM模型识别效果好,泛化能力强,进行数据增强后识别正确率得到了进一步的提升。Among EEG signal analysis methods,the time-frequency analysis method comprehensively considers the resolution of both the time and frequency of the signal,and improves the shortcomings of the simple time-domain or frequency-domain analysis method.In this experiment,S transform is used instead of STFT to transform left and right hand motor imagery EEG signals into two-dimensional time-frequency images,and then a CNN-ELM model is constructed for classification.When faced with the training data of small samples,the ability of the model is limited,so a data augmentation method is proposed.The generation of time-frequency images by conditional image synthesis with auxiliary classifier GANs(ACGAN)effectively enriched the number of training samples.The experimental results show that the CNN-ELM model has good recognition effect and strong generalization ability,and the recognition accuracy is further improved after data augmentation.

关 键 词：运动想象 脑电信号 S变换 卷积神经网络 极限学习机 数据增强 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]