基于小波变换和BP神经网络的视觉诱发电位识别  

Recognition of Visual Evoked Potential Based on Wavelet and BP Neural Network

作　　者：肖贵贤[1,2] 陈亚光[1] 官金安[1] 肖丹丹[1] 

机构地区：[1]中南民族大学电子信息工程学院,湖北430074 [2]湖北黄石理工学院,湖北435003

出　　处：《现代科学仪器》2008年第6期41-43,共3页Modern Scientific Instruments

基　　金：国家自然科学基金(30370393)

摘　　要：结合小波变换和误差逆传播(Error Back Propagation,BP)神经网络对视觉诱发脑电信号(visual evoked potential,VEP)进行分类而产生脑机接口控制信号。利用一维离散小波变换提取强噪声背景下的低频微弱脑电信号,获取特征向量输入BP神经网络进行事件相关电位模式识别。实验表明,小波变换特征向量提取方法能有效地实现信号的去噪、降维和特征提取,BP神经网络能比较准确地从VEP中识别出事件相关电位,进行10次测试的平均识别正确率为99.375%,有利于产生脑机接口控制信号。The classification of visual evoked potential（VEP） with wavelet transform and BP neural network was used to obtain brain-computer interface （BCI） control signal.One-dimensional discrete wavelet transform （DWT） method was used to extract the low-frequency weak VEP signal from strong background noise.The extracted feature vectors were input to the BP neural network to achieve the recognition of event related potential（ERP）.Experiments showed that the feature extraction in wavelet transformation could effectively extract VEP features,reduce noise and decrease the dimensionality,and the BP neural network could recognize ERP from VEP signal correctly.The average correct identification rate for 10 tests was 99.375%.It is useful to produce BCI control signal.

关 键 词：视觉诱发电位(VEP) 脑机接口 小波变换 BP神经网络 

分 类 号：R318[医药卫生—生物医学工程] TP18[医药卫生—基础医学]