基于神经网络的脚踝关节训练机器人设计  

Design of ankle joint training robot based on neural network

作　　者：郭燕婷 董颖敏 李景华 GUO Yanting;DONG Yingmin;LI Jinghua(Guangzhou Huali Technology Vocational College,Guangzhou,510000,China)

机构地区：[1]广州华立科技职业学院,广州510000

出　　处：《自动化与仪器仪表》2024年第9期268-271,共4页Automation & Instrumentation

基　　金：中国民办教育协会2023年度规划课题《高职康复专业传统康复课程的教学研究》(CANFZG23444)。

摘　　要：为进一步提升患者进行康复训练的体验效果,提出一种基于GBDT(Gradient Boosting Decision Tree,梯度提升树)算法和BP神经网络的脚踝关节康复训练机器人控制方法。其中,以GBDT算法作为患者运动动作模式的分类识别方法,以BP神经网络作为连续脚踝关节角度变化估计方法。实验结果表明,与其他分类识别算法相比,GBDT算法具有更高的识别精度和更快的识别效率,更加适用于动作模式的分类识别;使用BP神经网络能够进行准确的连续脚踝关节角度变化估计,同时还能够对角度的变化趋势进行良好拟合。综上,构建的脚踝关节康复训练机器人控制方法性能良好,可有效提升患者的使用体验,将其应用于实际康复训练场景中能够帮助患者进行更好的康复训练,可行性较高。To further enhance the experience of rehabilitation training for patients,a control method for ankle joint rehabilitation training robot based on GBDT(Gradient Boosting Decision Tree)algorithm and BP neural network is proposed.Among them,the GBDT algorithm is used as the classification and recognition method for patient movement patterns,and the BP neural network is used as the estimation method for continuous ankle joint angle changes.The experimental results show that compared with other classification and recognition algorithms,the GBDT algorithm has higher recognition accuracy and faster recognition efficiency,making it more suitable for the classification and recognition of action patterns;The use of BP neural network can accurately estimate the continuous changes in ankle joint angle,and can also fit the trend of angle changes well.In summary,the control method of the ankle joint rehabilitation training robot constructed has good performance and can effectively improve the user experience of patients.Applying it to actual rehabilitation training scenarios can help patients perform better rehabilitation training,with high feasibility.

关 键 词：康复训练 运动控制 GBDT BP神经网络 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]