A Transformer-Based Neural Network for Gait Prediction in Lower Limb Exoskeleton Robots Using Plantar Force

Lower limb exoskeleton robots have shown significant research value due to their capabilities of providing assistance to wearers and improving physical motion functions. As a type of robotic technology, wearable robots are directly in contact with the wearer's limbs during operation, necessitating a high level of human-robot collaboration to ensure safety and efficacy. Furthermore, gait prediction for the wearer, which helps to compensate for sensor delays and provide references for controller design, is crucial for improving the the human-robot collaboration capability. For gait prediction, the plantar force intrinsically reflects crucial gait patterns regardless of individual differences. To be exact, the plantar force encompasses a doubled three-axis force, which varies over time concerning the two feet, which also reflects the gait patterns indistinctly. In this paper, we developed a transformer-based neural network (TFSformer) comprising convolution and variational mode decomposition (VMD) to predict bilateral hip and knee joint angles utilizing the plantar pressure. Given the distinct information contained in the temporal and the force-space dimensions of plantar pressure, the encoder uses 1D convolution to obtain the integrated features in the two dimensions. As for the decoder, it utilizes a multi-channel attention mechanism to simultaneously focus on both dimensions and a deep multi-channel attention structure to reduce the computational and memory consumption. Furthermore, VMD is applied to networks to better distinguish the trends and changes in data. The model is trained and tested on a self-constructed dataset that consists of data from 35 volunteers. The experimental results show that FTSformer reduces the mean absolute error (MAE) up to 10.83%, 15.04% and 8.05% and the mean squared error (MSE) by 20.40%, 29.90% and 12.60% compared to the CNN model, the transformer model and the CNN transformer model, respectively.

Human Gait Modeling, Prediction and Classification for Level Walking Using Harmonic Models Derived from a Single Thigh-Mounted IMU

The majority of human gait modeling is based on hip, foot or thigh acceleration. The regeneration accuracy of these modeling approaches is not very high. This paper presents a harmonic approach to modeling human gait during level walking based on gyroscopic signals for a single thigh-mounted Inertial Measurement Unit (IMU) and the flexion–extension derived from a single thigh-mounted IMU. The thigh angle can be modeled with five significant harmonics, with a regeneration accuracy of over 0.999 correlation and less than 0.5° RMSE per stride cycle. Comparable regeneration accuracies can be achieved with nine significant harmonics for the gyro signal. The fundamental frequency of the harmonic model can be estimated using the stride time, with an error level of 0.0479% (±0.0029%). Six commonly observed stride patterns, and harmonic models of thigh angle and gyro signal for those stride patterns, are presented in this paper. These harmonic models can be used to predict or classify the strides of walking trials, and the results are presented herein. Harmonic models may also be used for activity recognition. It has shown that human gait in level walking can be modeled with a harmonic model of thigh angle or gyro signal, using a single thigh-mounted IMU, to higher accuracies than existing techniques.

Continuous Myoelectric Prediction of Future Ankle Angle and Moment Across Ambulation Conditions and Their Transitions

A hallmark of human locomotion is that it continuously adapts to changes in the environment and predictively adjusts to changes in the terrain, both of which are major challenges to lower limb amputees due to the limitations in prostheses and control algorithms. Here, the ability of a single-network nonlinear autoregressive model to continuously predict future ankle kinematics and kinetics simultaneously across ambulation conditions using lower limb surface electromyography (EMG) signals was examined. Ankle plantarflexor and dorsiflexor EMG from ten healthy young adults were mapped to normal ranges of ankle angle and ankle moment during level overground walking, stair ascent, and stair descent, including transitions between terrains (i.e., transitions to/from staircase). Prediction performance was characterized as a function of the time between current EMG/angle/moment inputs and future angle/moment model predictions (prediction interval), the number of past EMG/angle/moment input values over time (sampling window), and the number of units in the network hidden layer that minimized error between experimentally measured values (targets) and model predictions of ankle angle and moment. Ankle angle and moment predictions were robust across ambulation conditions with root mean squared errors less than 1° and 0.04 Nm/kg, respectively, and cross-correlations (R²) greater than 0.99 for prediction intervals of 58 ms. Model predictions at critical points of trip-related fall risk fell within the variability of the ankle angle and moment targets (Benjamini-Hochberg adjusted p > 0.065). EMG contribution to ankle angle and moment predictions occurred consistently across ambulation conditions and model outputs. EMG signals had the greatest impact on noncyclic regions of gait such as double limb support, transitions between terrains, and around plantarflexion and moment peaks. The use of natural muscle activation patterns to continuously predict variations in normal gait and the model's predictive capabilities to counteract electromechanical inherent delays suggest that this approach could provide robust and intuitive user-driven real-time control of a wide variety of lower limb robotic devices, including active powered ankle-foot prostheses.

Prediction of human gait activities using wearable sensors

This paper aims to enhance the accuracy of human gait prediction using machine learning algorithms. Three classifiers are used in this paper: XGBoost, Random Forest, and SVM. A predefined dataset is used for feature extraction and classification. Gait prediction is determined during several locomotion activities: sitting (S), level walking (LW), ramp ascend (RA), ramp descend (RD), stair ascend (SA), stair descend (SD), and standing (ST). The results are gained for steady-state (SS) and overall (full) gait cycle. Two sets of sensors are used. The first set uses inertial measurement units only. The second set uses inertial measurement units, electromyography, and electro-goniometers. The comparison is based on prediction accuracy and prediction time. In addition, a comparison between the prediction times of XGBoost with CPU and GPU is introduced due to the easiness of using XGBoost with GPU. The results of this paper can help to choose a classifier for gait prediction that can obtain acceptable accuracy with fewer types of sensors.