引入关联维的表面肌电信号的特征提取

考虑到表面肌电信号的非平稳特性,本研究在传统特征提取的基础上,又引入了非线性动力学中的关联维,通过小波系数的标准差和关联维重新构造特征向量,将其送入自组织映射网络对拇指弯曲、食指弯曲和无名指弯曲三种手势动作进行分类识别。结果表明:引入关联维的特征提取方法其识别正确率明显优于传统的小波变换的方法。可见,关联维作为一种新的特征参数,为肌电信号的特征提取提供了新的思路。

Feature Extraction of Surface Electromyography by Introducing Correlative Dimension

Considering the non-stationary character of surface EMG,the correlative dimension of nonlinear dynamics method of feature extraction was introduced.Through restructuring the feature vector by the correlative dimension and the standard error of the coefficients of wavelet transformation,three types of finger movement were identified employing self-organizing map.The result showed that the proposed method in this paper was much better than the traditional method of wavelet transformation.Obviously,as a new feature,the correlative dimension provides a new method for the feature extraction.