【摘要】 支持向量机 (SVM)是一种新的机器学习机制。研究了基于支持向量机的控制假手表面肌电识别方法和性能 ,并与反向传播 (BP)神经网络分类器进行了比较。分类的六种手腕部动作分别是腕内旋、腕外旋、展拳、握拳、肘部外旋、肘部内旋。利用“一对一”的分类策略和二叉树构造多类SVM分类器。核函数分别采用多项式和径向基函数。实验结果表明SVM可以有效地对表面动作肌电进行分类 ,SVM分类准确率普遍优于传统的BP神经网络 ,且具有良好的泛化推广能力。不同的核函数对分类准确率影响较小更多还原

【Abstract】 Support vector machine (SVM) is a new mechanism of machine learning. A Study on classifying surface electromyography (EMG) using SVM is presented for controlling prostheses.Moreover, a comparison study classification accuracy of EMG using SVM and back-propagation (BP) algorithm is given. The 6 motions to be classified are wrist flexion, wrist extension, hand close, hand open, ulnar deviation, and radial deviation.“One versus one” classification strategy and a binary tree structure multi-classes SVM classifier are used.The kernel functions are generated to use polynomials and radial basic functions. Results obtained show that the SVM could classify surface motion EMG more effectively and accurately than traditional BP neural network and it could be used more widely. Different kernel functions do not affect classification accuracy significantly.更多还原