基于SOA-ELM的手部动作识别方法实验研究  

Experimental Study of Hand Motion Recognition Method Based on SOA-ELM

作　　者：孟瑞[1] 岑豫皖[1] 王璐[1] 葛科铎 朱兴江 MENG Rui;CEN Yuwan;WANG Lu;GE Keduo;ZHU Xingjiang(School of Mechanical Engineering,Auhui University of Technology,Ma’anshan 243032,China;National Machine Tool Quality Supervision and Inspection Center,Ma’anshan 243000,China)

机构地区：[1]安徽工业大学机械工程学院,安徽马鞍山243032 [2]国家机床产品质量监督检验中心,安徽马鞍山243000

出　　处：《安徽工业大学学报（自然科学版）》2017年第3期264-268,共5页Journal of Anhui University of Technology（Natural Science）

基　　金：国家自然科学基金项目(51605005);安徽省自然科学基金项目(1708085MF164);安徽省高等学校省级自然科学研究重点项目(KJ2016A813)

摘　　要：为提高人体手部动作识别率,利用搜寻者优化算法(SOA)的全局搜索最优解原理和极限学习机(ELM)处理非线性关系的能力,提出一种基于搜寻者优化算法优化极限学习机(SOA-ELM)的手部动作识别方法。首先,采集内翻、外翻、握拳、展拳等4种手部动作的表面肌电信号,提取4种表面肌电信号的积分肌电值和均方根值,将其作为特征值;然后,利用这些特征值对ELM进行训练,采用SOA搜寻ELM的最优输入层权值和隐含层节点阈值;最后,采用经SOA优化的ELM对4种手部动作进行识别。实验结果表明,SOA-ELM比粒子群优化极限学习机(PSO-ELM)能更有效地对4种手部动作进行识别。To improve the accuracy of human hand motion recognition,a hand motion recognition method based on seeker optimization algorithm and extreme learning machine(SOA-ELM)was proposed,which took the advantage of the principle of SOAto search the global optimal solution and the ability of ELM to deal with nonlinear relation.Firstly,surface electromyography of 4 kinds of hand motions,such as wrist down,wrist up,hand grasps,hand extension were acquired,and the integral value and root mean square value of myoelectricity were extracted from 4 kinds of surface electromyography,which were taken as the feature values.Secondly,the feature values were used to train ELM,and the seeker optimization algorithm was used to find the optimal input layer weight and hidden layer threshold of ELM.Finally,4 kinds of hand motions were identified by SOA-ELM.Experimental results show that SOA-ELM is more effective than particle swarm optimization-extreme learning machine(PSO-ELM)for recognition of 4 kinds of hand motions.

关 键 词：表面肌电信号 搜寻者优化算法 极限学习机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]