| 佘青山,陈希豪,席旭刚,张启忠.基于DTCWT和CSP的脑电信号特征提取[J].,2016,56(1):70-76 |
| 基于DTCWT和CSP的脑电信号特征提取 |
| Feature extraction of EEG based on DTCWT and CSP |
| |
| DOI：10.7511/dllgxb201601011 |
| 中文关键词: 脑-机接口 运动想象 双树复小波变换 共空间模式 |
| 英文关键词: brain-computer interface (BCI) motor imagery dual-tree complex wavelet transform common spatial pattern (CSP) |
| 基金项目:国家自然科学基金资助项目(61201302)；国家留学基金资助项目(201308330297)；浙江省自然科学基金资助项目(LY15F010009)；浙江省国际科技合作项目(2013C24016). |
|
| 摘要点击次数: 1637 |
| 全文下载次数: 2267 |
| 中文摘要: |
| 针对运动想象脑电信号中存在很多与运动想象无关的频率信号和共空间模式特征提取方法缺少频率信息处理的问题，提出了一种双树复小波变换与共空间模式相结合的特征提取方法．该方法首先选取C3、Cz、C4 3个通道的脑电信号进行上采样，并利用双树复小波变换多尺度分解，获取适当的频段并在相应尺度下进行信号重构；然后将各频段的三通道重构信号联合输入到空间滤波器以得到所需的六维特征向量；最后利用支持向量机来完成两类运动想象任务的分类．采用BCI Competition Ⅳ提供的Dataset 1数据进行实验验证，与CSP、FBCSP、WPD-CSP方法进行比较，7名受试者的训练数据平均分类正确率可达到96.0%，测试数据平均分类正确率达到86.7%．实验结果表明了所提方法的有效性． |
| 英文摘要: |
| Due to the facts that there are many irrelevant frequency components in motor imagery electroencephalography (EEG), and the common spatial pattern (CSP) feature extraction method is lack of frequency information processing, a feature extraction method combining dual-tree complex wavelet transform (DTCWT) with CSP is presented. Firstly, three-channel EEG signals from C3, Cz and C4 are selected to be up-sampled, then the dual-tree complex wavelet transform is used to perform multi-scale decomposition to obtain appropriate bands and reconstruct the signals at the corresponding scale. Secondly, all bands of three-channel reconstructed signals are combined and inputted to the spatial filter, resulting in a 6-dimensional feature vector. Finally, support vector machines (SVM) is employed to classify two kinds of motor imagery tasks. Compared with the CSP, FBCSP and WPD-CSP methods, the proposed method is validated on Dataset 1 provided by BCI Competition Ⅳ, and its average classification accuracy of training data reaches 96.0% and that of testing data is 86.7% in seven subjects. Experimental results have demonstrated the effectiveness of the proposed method. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|