基于1维神经网络与快速傅里叶变换的SSVEP分类方法  

SSVEP classification method based on one-dimensional neural network and fast Fourier transform

作　　者：董延华[1] 景晨阳 王大东[1] DONG Yan-hua;JING Chen-yang;WANG Da-dong(College of Mathematics and Computer,Jilin Normal University,Siping 136000,China)

机构地区：[1]吉林师范大学数学与计算机学院,吉林四平136000

出　　处：《吉林师范大学学报(自然科学版)》2025年第1期117-124,共8页Journal of Jilin Normal University:Natural Science Edition

基　　金：中国高校产学研创新基金-新一代信息技术创新项目(2022IT096)。

摘　　要：提出了基于1维卷积神经网络和快速傅里叶变换(FFT-1DCNN)的稳态视觉诱发电位(Steady-State Visual Evoked Potential,SSVEP)信号分类方法,以满足非侵入式脑机接口(BCI)系统对于跨用户分类的需求.使用1维卷积神经网络捕捉局部依赖关系和模式,并使用快速傅里叶变换将多通道1维时域信号转换为2维频谱.实验采用低成本开销的8通道脑电信号放大器,以及特制的脑电头带以采集6名受试者的脑电数据,通过数据预处理以确保数据质量,并以自创数据集为基础,在用户依赖和用户独立两种训练情境中验证了所给算法的有效性.该方法在用户依赖情境中比TRCA算法准确率高3.14%,在用户独立情境中比CCA算法、FBCCA算法和C-CNN算法准确率分别高10.29%、4.72%和2.26%,证明了该方法的优越性.The steady-state visual evoked potential(SSVEP)is a type of brain signal with distinct characteristics,induced by exogenous visual stimuli,and has advantages such as quick training and low data requirements.However,the recognition and classification of SSVEP remains a challenge.This paper discusses an SSVEP signal classification method based on a one-dimensional convolutional neural network and fast fourier transform(FFT-1DCNN)to meet the needs of non-invasive brain-computer interface(BCI)systems for cross-user classification.The paper utilizes a one-dimensional convolutional neural network to capture local dependencies and patterns,and employs fast Fourier transform to convert multi-channel one-dimensional time-domain signals into two-dimensional spectra.This study uses a cost-effective 8-channel EEG signal amplifier and a specially made EEG headband to collect EEG data from six subjects,ensuring data quality through data preprocessing,and validates the effectiveness of the proposed algorithm based on a proprietary dataset in both user-dependent and user-independent training scenarios.In scenarios where users are dependent,this method surpasses the accuracy of the TRCA algorithm by 3.14%.In user-independent contexts,it outperforms the CCA,FBCCA,and C-CNN algorithms by 10.29%,4.72%,and 2.26%,respectively,proving its superior effectiveness.

关 键 词：脑机接口 稳态视觉诱发电位 卷积神经网络 傅里叶变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]