基于文化算法的表面肌电信号特征选择

Feature selection for surface electromyography signal using cultural algorithm

为了提高假肢控制系统肌电信号的分类准确率,提出一种新的基于文化算法的特征选择方法,通过该方法选择出最佳特征向量,然后用线性分类器检验其分类性能。利用表面差分电极从人体上肢四块肌肉采集四通道的肌电信号,对十个健康受试者进行八个动作的肌电信号模式分类实验,并同时用标准遗传算法来与文化算法作比较。实验结果表明,文化算法与遗传算法相比,特征维数更小,分类准确度更高。

To improve classification accuracy of the surface electromyography （sEMG）-based prosthesis, this paper proposed a new way to select feature based on cultural algorithm（CA） and used here. It tested its classification performance with linear discrimina analysis （LDA）. The method used surface differential electrodes to acquire four EMG signals from human body＇ s upper limbs. Ten healthy subjects participated in the experiment of classification of eight hand motion＇ s sEMG signals. Test results show that the algorithm can get a good result of classification. Compared with the standard genetic algorithm （ GA）, it has better search performance.