摘要
为了实现对智能下肢假肢进行有效控制,下肢步态(包括平地行走、上下楼梯和上下坡等)的有效识别是关键。先从提取不同步态下的特征值入手,利用平均影响值(MIV)来实现变量的筛选,并针对膝上截肢者的特点确定了6个特征值,分别为髋关节角度的最大值、支撑前期均值、支撑中期均值、支撑中期标准差、摆动期标准差(即Mh、ISh、MSh、SWh、MSV、SWV),然后利用概率神经网络(PNN)对本实验系统的5种步态进行准确识别,并与BP神经网络(BPNN)识别步态进行比较,试验结果表明将特征值用平均影响值方法筛选后,用概率神经网络进行步态识别,具有较好的识别率和识别速度,其识别率与BP神经网络相比提高了10%以上,验证了该方法的有效性和可行性。
In order to effectively control the intelligent lower limb prosthesis, the key is to effectively identify the gaits of the lower limb, including walk, up and down stairs or slopes, etc. Firstly, extract characteristic values of different gaits, screen the values with the mean impact value, and identify six feature values in accordance with the characteristics of the knee amputees, including the maximum value of hip joint angle, mean value in Initial Stance, mean value in middle stance, standard deviation in middle stance, standard deviation in swing phase of hip （ Mh, ISh, MSh, SWh, MSV, SWV）. Secondly, use the probabilistic neural network （PNN） to recognize them, which can accurately identify the five kinds of gaits in this experiment, and compare with BP neural network. The results show that the method, which uses the means of mean impact value to screen eigenvalues and recognizes gait by probabilistic neural network, has good recognition rate and recognition speed. The recognition rate which compared to BP neural network, increases by more than 10%, thereby verifying the effectiveness and feasibility of this methed.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2015年第2期181-185,共5页
Journal of Harbin Engineering University
基金
国家自然科学基金资助项目(61174009)
关键词
智能假肢
步态识别
平均影响值
概率神经网络
特征筛选
intelligent artificial limb
gait recognition
mean impact value
probabilistic neural network
feature selection
