Subject-Independent, Biological Hip Moment Estimation During Multimodal Overground Ambulation Using Deep Learning
Subject-Independent, Biological Hip Moment Estimation During Multimodal Overground Ambulation Using Deep Learning
复制标题
DOI：
10.1109/tmrb.2022.3144025
复制
发表时间：
2022-02-01
期刊：
影响因子：
--
通讯作者：
Young, Aaron J.
中科院分区：
文献类型：
--
作者：
Molinaro, Dean D.;Kang, Inseung;Young, Aaron J.
使用可穿戴传感器估计生物关节力矩可以实现实验室外生物力学分析和外骨骼，从而在日常生活中提供帮助。为了实现这些可能性，本研究引入了一种使用时间卷积网络 (TCN) 的独立于受试者的髋力矩估计器，并验证了其在多模式行走过程中的性能和通用性。当以全连接神经网络、长期短期记忆网络和基线方法（即使用基于移动模式和步态阶段的受试者平均力矩曲线）为基准时，使用来自 16 名参与者在平地、坡道和楼梯上行走的电测角仪和模拟 IMU 数据来评估我们的方法。此外，我们的方法的普遍性是通过对模型训练期间保留的地面坡度、楼梯高度和步态转换进行测试来评估的。 TCN 在保留数据方面优于基准方法 (p < 0.05)，稳态行走期间的平均 RMSE 为 0.131 +/- 0.018 Nm/kg，R-2 为 0.880 +/- 0.030。在 20 个留一法坡度和楼梯高度条件下进行测试时，TCN 仅在最陡（+18 度）坡度上显着增加 RMSE（p < 0.05）。最后，在模式转换期间，TCN RMSE 和 R-2 分别为 0.152 +/- 0.027 Nm/kg 和 0.786 +/- 0.055。因此，我们的方法使用来自三个可穿戴传感器的数据准确地估计髋关节力矩并推广到不可见的步态环境。
Estimating biological joint moments using wearable sensors could enable out-of-lab biomechanical analyses and exoskeletons that assist throughout daily life. To realize these possibilities, this study introduced a subject-independent hip moment estimator using a temporal convolutional network (TCN) and validated its performance and generalizability during multimodal ambulation. Electrogoniometer and simulated IMU data from sixteen participants walking on level ground, ramps and stairs were used to evaluate our approach when benchmarked against a fully-connected neural network, a long short-term memory network, and a baseline method (i.e., using subject-average moment curves based on ambulation mode and gait phase). Additionally, the generalizability of our approach was evaluated by testing on ground slopes, stair heights, and gait transitions withheld during model training. The TCN outperformed the benchmark approaches on the hold-out data (p < 0.05), with an average RMSE of 0.131 +/- 0.018 Nm/kg and R-2 of 0.880 +/- 0.030 during steady-state ambulation. When tested on the 20 leave-one-out slope and stair height conditions, the TCN significantly increased RMSE only on the steepest (+18 degrees) incline (p < 0.05). Finally, the TCN RMSE and R-2 was 0.152 +/- 0.027 Nm/kg and 0.786 +/- 0.055, respectively, during mode transitions. Thus, our approach accurately estimated hip moments and generalized to unseen gait contexts using data from three wearable sensors.