摘要
为了去除在脑电信号采集过程中受到的干扰,在传统方法的基础之上,提出了一种基于卷积神经网络(CNN)脑电信号伪迹检测与去除的方法。该方法通过CNN模型对脑电信号电压幅值计算后的特征进行提取,完成Softmax分类器对脑电信号的检测分类。采用EEMD算法将含噪脑电信号分解为若干个本征模式函数IMF分量,通过Hilbert特征法提取出噪声占主导的高频IMF分量,再由FastICA的方法将剩余信号分离,达到眼电伪迹的去除。实验表明,CNN方法检测准确率高达80%以上,CNN与EEMD的结合提高了脑电信号伪迹去除的有效性。
In order to eliminate the interference in the process of EEG acquisition,based on traditional methods,a method of detecting and removing EEG artifacts based on convolutional neural network(CNN)is proposed in this paper.Through extracting the characteristics after calculating the voltage amplitude of the EEG by CNN model,the classification of EEG detection is completed by Softmax classifier.The EEMD algorithm is used to decompose the noisy EEG into a number of eigen mode function IMF components,and extract the high-frequency IMF components of the dominant noisy EEG by Hilbert feature method.Then the residual signal is separated by the FastICA method to remove the artifact of the eye.The experiment shows that the detection accuracy of CNN method is up to 80%,and the combination of CNN and EEMD improves the effectiveness of the artifacts removal of the EEG.
作者
张晨洁
王爽
郭滨
白雪梅
ZHANG Chenjie;WANG Shuang;GUO Bin;BAI Xuemei(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022)
出处
《长春理工大学学报(自然科学版)》
2018年第2期119-123,128,共6页
Journal of Changchun University of Science and Technology(Natural Science Edition)
基金
吉林省科技厅自然科学基金项目(20150101013JC)
关键词
卷积神经网络
总体经验模态分解
希尔伯特变换
脑电信号去噪
convolutional neural network(CNN)
ensemble empirical mode function(EEMD)
Hilbert transformation
brain electrical signal to noise