摘要
为了更好地识别人手动作的肌电信号,采用基于小波包分解与主成分分析结合的特征提取方法,并利用粒子群优化Elman神经网络的模式分类方法。选择“db”系小波对肌电信号进行多尺度分解,并结合主成分分析法,选用累计贡献率大于98.6%的十个主成分作为特征向量,输入优化神经网络进行网络训练,实现对人手抓取动作的模式识别。实验结果表明,与传统神经网络仿真结果对比,采用粒子群算法优化Elman神经网络不仅能提高系统稳定性问题,而且能提高人手动作分类识别率,验证了该方法是一种可行的人手动作分类识别方法。
In order to better identify sEMG signals ofhuman hand motion,a feature extraction method based on wavelet packet decomposition and principal component analysis is adopted.And the pattern classification method of Elman neural network is optimized by particle swarm optimization.The researcher selected“db”system wavelet to multi-scale decomposition of sEMG signal combined with principal component analysis method.Aiming at ten principal components with cumulative contribution rate greater than 98.6%as feature vector,the researcher input optimized neural network for network training to realize the pattern identification of human hand grasp.The experimental results show that compared with the traditional neural network simulation results,using particle swarm optimization algorithm to optimize Elman neural network can not only solve the stability problem of neural network,but also improves the classification rate of human motion classification,and therefore verifies that the method is a feasible classification identification method of human hand movement.
作者
雷华勤
Lei Huaqin(Fuzhou Vocational Education Training Center,Fuzhou 350009,Fujian;Fuzhou University,Fuzhou 350108,Fujian)
出处
《武汉工程职业技术学院学报》
2019年第4期13-17,共5页
Journal of Wuhan Engineering Institute
关键词
表面肌电信号
主成分分析
小波包分解
ELMAN神经网络
粒子群优化算法
SEMG
PSO
surface electronomyography signal
principal component analysis
wavelet packet decomposition
Elman neural network
particle swarm optimization algorithm
sEMG
PSO
