基于小波方差的ECoG分类

• 作 者: 颜世玉 刘冲 王宏 赵海滨
• 机 构: 东北大学机械工程与自动化学院,
• 出 处: 生物医学工程学杂志
• 发表日期: 2013年03期
• 关键词: 小波方差 小波变换 皮层脑电图 脑-机接口 ECoG 交叉验证 特征提取 分类正确率 classify 线性分类

针对典型的基于皮层脑电图(ECoG)的脑-机接口(BCI)设计,被试任务为想象左手小指和想象舌头运动,提出了采用小波方差的特征提取方法。首先在小波变换的基础上,提出小波方差的计算方法及其意义,并以此作为特征,从64导联中获取特征较为明显的6个导联进行分析;然后对脑电(EEG)数据进行三层小波分解,根据ERD/ERS现象,提取包含Mu节律和Beta节律的小波系数方差作为特征,采用交叉验证的方法,利用classify进行线性分类。离线分析表明,对训练集和测试集的分类正确率达到90.24%和93.77%,小波方差作为BCI研究中特征提取的方法具有更加简单和有效的特性。

Aiming at the typical ECoG-based brain-computer interface (BCI) design, the task of the subject proposed a feature extraction method using wavelet variance to visualize the movement of the left little finger and the imagined tongue. Firstly, based on the wavelet transform, the method of calculating the variance of the wavelet and its significance are proposed. Based on this, six leads with obvious characteristics are obtained from the 64 leads. Then the EEG data are analyzed Based on the ERD / ERS phenomenon, the variance of wavelet coefficients including Mu rhythm and Beta rhythm is extracted as the feature, and the cross-validation method is used to classify the samples by classify. Off-line analysis shows that the classification accuracy of training set and test set is 90.24% and 93.77%, respectively. The wavelet variance is more simple and effective as a feature extraction method in BCI.