基于角速度的膝关节屈伸角度实时估计方法

Real-time estimation method of knee joint flexion and extension angle based on angular velocity

穿戴式膝关节屈伸角度的实时估计对康复评估及外骨骼康复机器人的实时控制有重要作用。利用穿戴在大腿和小腿上的两个角速度传感器实现膝关节角度实时估计。首先离线通过遗传算法优化计算膝关节模型的关节轴,然后在线将这两个角速度传感器数据投影到关节轴上,确定出膝关节屈伸角度变化的角速度差,对角速度差进行积分可得膝关节屈伸角度。同时,利用小腿上的角速度传感器实时判断脚跟着地点时刻,在该时刻对角度进行重置来消除积分漂移。募集5个健康人和2个膝骨性关节炎患者进行实验。结果表明,小腿上的角速度传感器能准确判断脚跟着地点,与压力传感器判断出的脚跟着地时间点相比较,在5个步态周期里,健康组的平均均方根误差为27.88±19.64 ms,患者组的平均均方根误差为54.60±7.21 ms,约占步态周期的3%,有较高精度,符合应用需求。本方法的膝关节屈伸角度估计方法和角度传感器的实际测量数据对比,健康组的平均均方根误差为2.86±0.53°,患者组测试的平均均方根误差为2.00±0.78°。改变大腿上传感器穿戴位置的实验结果为穿戴位置不影响模型估计的角度。本方法操作方便,有效消除积分漂移,实现对膝关节屈伸角度的实时准确估计。

The real-time estimation of knee flexion and extension angle using wearable equipment plays an important role in rehabilitation evaluation and real-time control of the exoskeleton rehabilitation robot. Two angular velocity sensors on the thigh and shank are used to realize real-time estimation of the knee joint angle. Firstly, the joint axis of knee joint model is optimized by genetic algorithm offline. Then, the data of these two angular velocity sensors are projected on the joint axis online to calculate the angular velocity relative difference of knee joint flexion and extension angle. The knee joint flexion and extension angle can be achieved by integrating the angular velocity relative difference. Meanwhile, the angular velocity sensor on the shank is used to detect heel-strike in real time, and the angle is reset at this time to eliminate the integral drift. Five healthy people and two patients with knee osteoarthritis are recruited for experiments. Results show that the angular velocity sensor on the lower leg can accurately determine the event of heel-strike. Compared with the time of heel-strike determined by the pressure sensor, the average root mean square error(RMSE) of the healthy group is 27.88±19.64 ms. And RMSE of the patient group is 54.60±7.21 ms, which accounts for about 3% of the gait cycle. It has high accuracy and can meet the application requirements. The average RMSE of the healthy group is 2.86±0.53° and RMSE of the patient group is 2.00±0.78°. Experimental results show that the wearable position does not affect the angle estimated by the model. This method is easy to operate. The integral drift can be effectively eliminated. The real-time accurate estimation of knee flexion and extension angle can be realized.