Kinematics, Speed, and Anthropometry-Based Ankle Joint Torque Estimation: A Deep Learning Regression Approach

Powered Assistive Devices (PADs) have been proposed to enable repetitive, user-oriented gait rehabilitation. They may include torque controllers that typically require reference joint trajectories determine the most suitable level of assistance. However, a robust approach able automatically estimate trajectories, namely ankle torque, while considering effects varying walking speed, body mass, and height on dynamics, is needed. This study evaluates accuracy generalization ability two Deep Learning (DL) regressors (Long-Short Term Memory Convolutional Neural Network (CNN)) generate by innovatively customizing them according speed (ranging from 1.0 4.0 km/h) users’ mass 1.51 1.83 m 52.0 83.7 kg, respectively). Furthermore, this hypothesizes DL can without resourcing electromyography signals. CNN was algorithm (Normalized Root Mean Square Error: 0.70 ± 0.06; Spearman Correlation: 0.89 0.03; Coefficient Determination: 0.91 0.03). No statistically significant differences were found in (p-value > 0.05) whether signals are included as inputs or not, enabling less obtrusive accurate setup for estimation.