摘要
在基于表面肌电信号的人机交互系统中,产生的肌肉疲劳降低了系统的稳定性。针对该问题,分析肌肉正常状态和疲劳状态下的肌电信号变化规律,提出一种改进的在线支持向量机增量训练算法。该算法在每次训练SVM(Support Vector Machine)模型时,计算各样本到分类超平面的距离,并以之为条件对不断更新的训练数据进行有条件的选择和遗忘,只留下最大距离1/2以内的数据。通过在线训练不断更新训练样本来获得新的SVM模型,用于适应肌肉疲劳过程中肌电信号的变化,同时防止多次在线训练过程中更新的样本改变训练集间初始边界。最后在智能轮椅上进行验证,实验结果表明:该算法有效减少了肌肉疲劳在人机交互系统中的影响,使得系统能够保持长时间稳定操作。
For the problem that the stability of surface Electromyograph （sEMG） based human -machine interface （HMI） declines as the muscle fatigue takes place, an improved incremental training algorithm for online support vector machine （SVM） is proposed. The novel method adjusts the model of SVM to adapt itself based on the changes of sEMG and the training data are conditionally selected and forgotten by the distance of samples to the classification hyperplane. The data are left which is less than half the maximum distance. Through online updating the training samples, a new SVM model is obtained to used to adapt to the changes of EMG, and prevents the change of initial boundary among training sets of sample during the online training process. Intelligent wheelchair experiment results show that the presented algorithm performs high modeling precision and training speed is increased. Furthermore, this method effectively overcomes the influence of muscle fatigue during longterm operating sEMG based HMI.
出处
《控制工程》
CSCD
北大核心
2014年第4期467-471,共5页
Control Engineering of China
基金
国家自然科学基金项目(60905066,51075420)
科技部国际合作项目(2010DFA12160)
重庆市科技攻关项目(CSTC,2010AA2055)
重庆市教委科学技术研究项目(KJ100516)
