摘要
为了提高人体下肢多运动模式识别的准确性,提出一种基于多源信息和粒子群优化算法-误差反向传播(PSO-BP)神经网络的识别方法.建立下肢多源信息采集系统,该系统由下肢表面肌电信号、髋关节角度、髋关节加速度组成.选择肌电信号偏度、峭度和功率谱比值为肌电信号特征,髋关节角度细分模式均值比为腿部角度信号特征,加速度标准差、能量峰值、两轴相关性系数为髋关节加速度特征.按照主成分分析(PCA)方法融合上述特征值,利用PSO-BP进行识别.实验结果表明:该方法识别率为95.75%,平均识别时间为1.234 8s.
An approach based on multi-source information and particle swarm optimization algorithm-back propagation（PSO-BP）neural network was proposed in order to improve the accuracy of human lower limb locomotion-mode identification.A multi-source information acquisition system was established,which was composed of lower limb surface electromyography signal（sEMG）,hip joint angle and hip joint acceleration.Specifically,skewness,kurtosis and power spectrum ratio were extracted from surface electromyography（sEMG）;the average ratio of hip angle of segmentation mode was extracted from gyroscope;standard deviation,peak of energy and correlation coefficient were extracted from accelerometer.Principal component analysis（PCA）was used to fuse these features.PSO-BP neural network was trained using an experimental database for locomotion-mode identification.The test results indicated that the locomotionmode identification rate was 95.75%,and the average identification time was 1.234 8s.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2015年第3期439-447,共9页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(61174009
61203323)
天津市自然科学基金资助项目(13JCQNJC03400)
关键词
多源信息
粒子群优化算法
主成分分析
下肢运动模式识别
BP神经网络
multi-source information
particle swarm optimization algorithm
principal component analysis
lower limb locomotion-mode identification
BP neural network
