基于CNN-CBAM-LSTM的稳态视觉诱发电位脑电信号识别方法  

A Recognition Method for Steady-State Visual Evoked Potential EEG Signals Based on CNN-CBAM-LSTM

作　　者：巩炫麟 陶庆[2] 苏娜[3] 马金旭 GONG Xuan-lin;TAO Qing;SU Na;MA Jin-xu(School of Intelligent Manufacturing Modern Industry(School of Mechanical Engineering),Xinjiang University,Urumqi 830017,China;College of Engineering,Xinjiang University,Urumqi 830017,China;The First Affiliated Hospital of Xinjiang Medical University,Urumqi 830054,China)

机构地区：[1]新疆大学智能制造现代产业学院(机械工程学院),乌鲁木齐830017 [2]新疆大学工程师学院,乌鲁木齐830017 [3]新疆医科大学第一附属医院,乌鲁木齐830054

出　　处：《科学技术与工程》2025年第10期4175-4182,共8页Science Technology and Engineering

基　　金：新疆维吾尔自治区“天山英才”科技创新领军人才项目(2023TSYCLJ0051)。

摘　　要：在使用传统方法处理稳态视诱发电位(steady-state visual evoked potentials,SSVEP)的脑电信号时,特征提取的准确性和充分性存在不足,影响信号的识别准确率。为此提出了一种基于卷积神经网路(convolutional neural networks,CNN)与卷积注意力机制模块(convolutional block attention module,CBAM)和长短时记忆网络(long short-term memory,LSTM)的信号分类识别方法。以CNN为基础框架,通过引入注意力机制对通道及空间特征进行充分提取,加入LSTM提高对时序特征的提取能力,实现对SSVEP信号的目标识别。实验结果显示,所提方法能够充分有效的提取各级特征且识别准确率较高,相比于典型相关分析方法(canonical correlation analysis,CCA)、CNN、CBAM-LSTM、CNN-CBAM识别准确率分别提高了5.3%、2.95%、2.27%、1.71%,可见该模型对SSVEP信号的分类识别有较好的效果。When traditional methods were used to evoked the potentials SSVEP(steady-state visual evoked potentials)EEG(electroencephalogram)signals,the accuracy and sufficiency of feature extraction were insufficient,which affected the recognition accuracy of signals.A novel approach was proposed which based on a CNN(convolutional neural network)integrated with a CBAM(convolutional block attention module)and a LSTM(long short-term memory network).By incorporating attention mechanisms,both channel and spatial features were effectively extracted within the CNN framework.Additionally,LSTM was introduced to enhance the extraction of temporal features,enabling accurate recognition of SSVEP signals.The experimental results show that the proposed method can effectively extract hierarchical features and achieves a high recognition accuracy.Compared to canonical correlation analysis(CCA),CNN,CBAM-LSTM,and CNN-CBAM,the proposed model improves the recognition accuracy by 5.3%,2.95%,2.27%,and 1.71%respectively.It can be seen that the model has a good performance in the classification and recognition of SSVEP signals.

关 键 词：稳态视觉诱发电位 卷积神经网络 卷积注意力机制模块 长短时记忆网络 目标识别 

分 类 号：TN911.7[电子电信—通信与信息系统]