SVM在表面肌电信号手部动作模式识别中的应用

Application of SVM in Handmotion Pattern Recognition Based on Surface Electromyography

肌电信号蕴含着丰富的神经肌肉运动信息,对肌电信号进行模式识别是实现肌电假肢控制的基础。特征提取是模式识别中关键的一环,模式识别效果与特征值的选取以及分类器的设计具有直接关系。支持向量机(SVM)是一种基于统计学习理论的机器学习方法。该方法可自动寻找出那些对分类有较好区别能力的支持向量,由此构造出的分类器对于小样本、非线性情况具有很好的分类效果。文章通过提取表面肌电信号小波变换系数绝对值的最大值与时频域其他特征值一起构成特征向量,设计了SVM分类器,对六种手部动作(握拳、伸掌、内旋、外旋、屈腕、伸腕)实现了分类,平均准确率达到97.5%,验证了SVM分类器对肌电信号动作模式识别的有效性。

EMG signals contained abundant neuromuscularmovement information,and pattern recognition of EMG signals was the basis for EMG prosthesis control.Feature extraction was a key link in pattern recognition.The effect of pattern recognition was directly related to the selection of feature values and the design of classifiers.Support vectormachine(SVM)was a machine learning method based on statistical learning theory.The method could automatically find out those support vectors that had good distinguishing ability for classification.The classifier constructed by this method had good classification effect for small samples and non-linear cases.In this paper,by extracting themaximum value of the absolute value of the wavelet transformed coefficient of surface electromyography signals and other eigen values in the time-frequency domain to form the eigenvector,one SVM classifier was designed to classify six kinds of handmovements(clenching,stretching palm,pronation,supination,wrist flexion and wrist extension),with an average accuracy rate of 97.5%,which verified the effectiveness of SVM classifier in electromyography signal action pattern recognition.