%0 Journal Article %T 基于传感器的智能膝关节矫形器力矩预测方法及系统
Sensor-Based Intelligent Knee Orthosis Moment Prediction Method and System %A 皇艳波 %A 徐彼得 %A 吴淏 %A 马寻君 %A 吴云成 %J Modeling and Simulation %P 514-524 %@ 2324-870X %D 2025 %I Hans Publishing %R 10.12677/mos.2025.144305 %X 膝关节骨性关节炎作为高致残性疾病，传统治疗手段受限于实验室评估的实时性与便捷性。本研究提出一种基于多源传感器的智能膝关节矫形器力矩预测系统，通过融合3D打印矫形器、惯性测量单元(IMU)及表面肌电传感器，实时监测膝关节内收力矩(KAM)与屈曲力矩(KFM)。通过构建融合一维卷积神经网络、双向长短期记忆网络及注意力机制(CNN-BiLSTM-Attention)的深度学习模型，以动态预测关节力矩。实验招募10名健康受试者，采集步行、跑步、上下楼梯及转向等六类日常活动的运动数据，利用逆动力学计算力矩作为基准。结果显示，KAM预测性能优于KFM：KAM平均相关系数为0.739 ± 0.094，均方根误差(RMSE)为0.172 ± 0.033 Nm/kg；KFM平均相关系数为0.720 ± 0.107，RMSE为0.215 ± 0.048 Nm/kg。研究进一步验证了传感器数据与云平台结合的长期康复管理可行性。该系统突破了传统实验室评估的局限，为个性化矫形器干预和智能康复设备的发展提供了创新解决方案。
Knee osteoarthritis, as a highly disabling disease, has traditional treatment methods limited by the real-time and convenience of laboratory assessments. This study proposes an intelligent knee orthosis torque prediction system based on multi-source sensors, which integrates 3D-printed orthoses, inertial measurement units (IMUs), and surface electromyography sensors to monitor knee adduction moment (KAM) and knee flexion moment (KFM) in real time. By constructing a deep learning model that combines one-dimensional convolutional neural networks, bidirectional long short-term memory networks, and attention mechanisms (CNN-BiLSTM-Attention), the system dynamically predicts joint torque. The experiment recruited 10 healthy subjects, collecting motion data from six types of daily activities including walking, running, ascending and descending stairs, and turning. Inverse dynamics were used to calculate the torque as a benchmark. The results show that the prediction performance for KAM is superior to that for KFM: the average correlation coefficient for KAM is 0.739 ± 0.094, with a root mean square error (RMSE) of 0.172 ± 0.033 Nm/kg; the average correlation coefficient for KFM is 0.720 ± 0.107, with an RMSE of 0.215 ± 0.048 Nm/kg. The study further validates the feasibility of long-term rehabilitation management by combining sensor data with a cloud platform. This system breaks through the limitations of traditional laboratory assessments and provides an innovative solution for personalized orthotic intervention and the development of intelligent rehabilitation equipment. %K 膝骨关节炎， %K 膝关节矫形器， %K 力矩预测， %K 传感器， %K 深度学习
Knee Osteoarthritis %K Knee Orthosis %K Moment Prediction %K Sensors %K Deep Learning %U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=112074