基于松散型WNN的运动想象脑电信号解码研究  

Decoding of motor imagery EEG signals based on loose WNN

作　　者：胡秀枋[1] 何爽 邹任玲[1] 张一凡 毛晨罡 黄鑫[1] 李丹[1] 曹立 HU Xiufang;HE Shuang;ZOU Renling;ZHANG Yifan;MAO Chengang;HUANG Xin;LI Dan;CAO Li(School of Health Science and Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China;Shanghai Sixth People's Hospital,Shanghai 200233,China)

机构地区：[1]上海理工大学健康科学与工程学院,上海200093 [2]上海市第六人民医院,上海200233

出　　处：《智能计算机与应用》2023年第11期239-243,共5页Intelligent Computer and Applications

基　　金：上海市科技创新行动计划产学研医合作领域项目(21S31906000);上海理工大学医工交叉项目(10-22-308-524)。

摘　　要：基于运动想象的脑机接口可以控制外部设备,在医疗康复领域中有着重要的临床意义。为了提高运动想象脑电信号的分类准确率,提出一种松散型小波神经网络,即双树复小波变换与神经网络分开进行计算再组合。利用DTCWT对预处理后的脑电信号进行分解,计算复小波系数的多个特征值并构建特征向量,将组合后的特征向量送入神经网络中进行分类识别。实验结果表明,该算法在BCI Competition IV的数据集2a上的平均准确率为76.03%。通过与不同分类器和现有方法的比较,验证了松散型小波神经网络的有效性。The brain computer interface based on motor imagery can control external devices and has important clinical significance in the field of medical rehabilitation.In order to improve the classification accuracy of motor imagery EEG signals,a loose wavelet neural network is proposed,namely,the dual-tree complex wavelet transform and neural network are computed separately and then combined.The pre-processed EEG signals are decomposed using DTCWT,multiple eigenvalues of the complex wavelet coefficients are calculated and feature vectors are constructed,and the combined feature vectors are fed into the neural network for classification and recognition.The experimental results show that the algorithm has an average accuracy of 76.03%on dataset 2a of BCI Competition IV.The effectiveness of the loose wavelet neural network is verified by comparing it with different classifiers and existing methods.

关 键 词：运动想象 松散型小波神经网络 双树复小波变换 脑机接口 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]