摘要
为更准确地生成用于被动康复训练的个性化参考步态,提出了一种结合随机森林(RF)和概率运动基元(ProMPs)的步态模式预测方法。首先使用RF根据个体身体特征参数和指定步速预测个体的步态特征,然后引入ProMPs学习正常人步态数据的概率分布并基于预测的步态特征进行步态模式重构。实验结果表明:与现有方法相比,RF-ProMPs预测的髋膝关节步态模式平均绝对误差(MAD)分别降低了32%和18%,该方法可用于生成下肢被动康复训练的参考轨迹。
To generate individualized reference gait pattern for passive rehabilitation training more accurately,a gait pattern prediction model combining random forest(RF)and probabilistic movement primitives(ProMPs)is proposed.Firstly,RF is used to predict the gait characteristics of individuals based on their physical characteristic parameters and specified gait speed.Then,ProMPs are introduced to learn the probability distribution of normal human gait data and reconstruct gait patterns based on the predicted gait characteristics.The experimental results show that compared with the existing methods,the mean absolute deviation(MAD)of the hip and knee gait patterns predicted by the RF-ProMPs model are reduced by 32%and 18%,respectively.The proposed method can be used to generate reference trajectories for lower limb passive rehabilitation training.
作者
曾春鸿
陆康
何志琴
吴钦木
ZENG Chunhong;LU Kang;HE Zhiqin;WU Qinmu(College of Electrical Engineering,Guizhou University,Guiyang 550000,China)
出处
《传感器与微系统》
北大核心
2026年第4期158-161,共4页
Transducer and Microsystem Technologies
基金
贵州省科技计划项目(黔科合支撑[2021]一般442,黔科合支撑[2023]一般179)。
