摘要
良好的下肢步态感知性能有助于提升助力型外骨骼机器人的助力效果。本文以足底压力分布为研究对象,基于足底生物力学分析搭建一种可穿戴式足底压力分布采集装置,分别采集平地行走、平地慢跑和坡路行走3个步态的足底压力数据,并基于多元线性回归法构建的地面反作用力预测模型获取整体足压,提出了一种基于整体足压和卷积神经网络(CNN)分类算法进行下肢步态识别的方法,并与支持向量机(SVM)和反向传播(BP)神经网络进行了对比分析。试验结果表明:该方法对于3种步态的平均识别率达到98.3%,具有较高的准确性,验证了使用CNN分类算法对下肢不同步态识别的可行性与有效性。
Good lower limb gait perception performance can help improving assistance effectiveness of exoskeleton robots.Taking plantar pressure distribution as research object,a wearable plantar pressure distribution collection device is built based on plantar biomechanical analysis,plantar pressure data of three different gaits,namely walking on flat ground,slow jogging on flat ground,and walking on slopes,are collected,respectively.Overall plantar pressures are obtained by constructing a ground reaction force prediction model based on multiple linear regression method.A method for lower limb gait perception based on the overall plantar pressure and convolutional neural network(CNN)classification algorithm is proposed.Comparative analysis are conducted on support vector machine(SVM)and back propagation(BP)neural network.The experimental results show that the proposed method achieves an average recognition rate of 98.3%on the three gaits,and has higher accuracy.The feasibility and effectiveness of using the CNN classification algorithm to identify different lower limb gaits are verified.
作者
颜兵兵
王强
宋佳宝
殷宝麟
胡春玉
YAN Bingbing;WANG Qiang;SONG Jiabao;YIN Baolin;HU Chunyu(College of Mechanical Engineering,Jiamusi University,Jiamusi 154007,China)
出处
《传感器与微系统》
CSCD
北大核心
2024年第2期143-147,共5页
Transducer and Microsystem Technologies
基金
高等教育本科教育教学改革研究重点委托项目(SJGZ20220123)
黑龙江省高等学校基本科研业务费科研项目(2021—KYYWF—0562)
佳木斯大学青年创新人才培养计划项目(JMSUQP2022003)
2022年度国家级大学生创新创业训练计划项目(202210222074)。
关键词
助力型外骨骼
足底压力
穿戴式采集装置
预测模型
步态识别
assistive exoskeletons
plantar pressure
wearable collection device
predictive model
gait recognition
