12 April 2022 in IEEE Transactions on Medical Robotics and Bionics
This study aims to develop a robust myoelectric control method for gait phase recognition in a lower-limb exoskeleton robot. In the proposed method, a metric learning-based temporal convolution network (ML-TCN) was utilized to extract discriminative features of the surface electromyography (sEMG) signals and recognize four gait phases: heel strike (HS), foot flat (FF), heel off (HO), and swing (SW). Vicon validated the effectiveness of the sEMG data-acquisition system. The proposed method acquires significantly more discriminative features than long short-term memory (LSTM) or common temporal convolutional network (TCN). The experimental results show that the proposed model has a higher prediction accuracy and stronger robustness against disturbances in complex terrain. Finally, under the complex terrain of level ground-ramp ascending-ramp descending walking, the proposed model’s accuracy of gait phase recognition is 96.22%, which is better than LSTM’s 91.20%. Noise disturbances of 10%, 20%, 30%, 40%, and 50% were added to the test set. Compared with LSTM, the resistance to disturbances of the proposed method increased by 8.15%, 8.79%, 9.67%, 10.6%, and 10.61%, respectively.
23 June 2022 in IEEE Robotics and Automation Letters
Lower-limb exoskeleton robots can support hemiplegic patients’ affected limbs and assist in their rehabilitation. In order to set effective control strategies, it is necessary to obtain the user’s motion intention accurately and timeously. These requirements pose many challenges. The surface electromyography (sEMG) signal has long been used for detecting a person’s locomotion mode. However, most traditional myoelectric motion recognition methods need to collect multi-channel sEMG signals (more than eight) and use feature engineering to obtain the desired accuracy. Additionally, traditional methods make decisions when the recognition probability is still uncertain, resulting in misclassification during continuous recognition. In order to overcome these limitations, a dual-purpose autoencoder-guided temporal convolution network (DA-TCN) is proposed, which is used as a motion detection model for a lower-limb exoskeleton robot based on sEMG. In the proposed method, we first obtain four-channel sEMG signals without any feature engineering. Next, the DA-TCN is used to obtain the discriminative deep features of the sEMG signals. Compared with separable features from common TCNs, discriminative features assist in obtaining a more discriminative twins-loss. Finally, the classification result is redetermined by the decoder, based on the reconstruction of deep features, and misclassifications are rejected based on the value of the twins-loss. The performance of the proposed method was evaluated by obtaining data from seven volunteer subjects in seven locomotion modes. The proposed method was able to effectively detect misclassifications and significantly improved the stability of continuous recognition.
9 January 2023 in IEEE Robotics and Automation Letters
Multi-stream signals are increasingly being used in robot-assisted rehabilitation training, where the timely and accurate prediction of a patient's motor intentions is frequently required to provide simultaneous and proportional control strategies. However, existing methods for motion intent prediction typically isolate signals from different modalities, resulting in a loss of signal traceability and interpretability. Therefore, we propose a multi-stream signal-fusion strategy based on knowledge tracing (MSKT). First, we collected surface electromyography, force myography, and vibroarthrography signals from the lower extremity and used different encoders to map these signals into a problem component, concept component, and identity token component. The process of human learning was then simulated using a knowledge-tracing strategy to reconstruct these components into new features. Finally, the reconstructed set of depth features was mapped to joint angles. We validated the performance of the proposed method using motion data from eight participants in different locomotion modes and compared the results to those of a temporal convolution network and long short-term memory networks. The results indicate that the MSKT achieves the lowest root-mean-squared error. This study demonstrates that the proposed knowledge-tracing-based strategy enables accurate continuous estimation for lower-limb exoskeleton rehabilitation robots.
12 March 2025 in IEEE Transactions on Medical Robotics and Bionics
In this study, we propose a robust myoelectric intention recognition framework to recognize human locomotion mode and detect irrelevant locomotion intention. The framework is integrated into the control system of the lower limb exoskeleton robot for experimental validation. Most conventional electromyography(EMG) intention detection methods aim to accurately detect the target motion intentions but ignore the possible effects of irrelevant intentions. In traditional action intention recognition strategies, most researchers did not consider entering irrelevant action intentions into the model during training. Therefore, when using a classification model, if irrelevant action intentions are input, the model will still recognize it as a type of target action intention. That can lead to incorrect recognition results, which will cause the robot to perform wrong movements and pose a safety risk to the wearer. To detect and reject irrelevant motion intentions, we first used the dual-purpose autoencoder-guided temporal convolution network (DA-TCN) to obtain discriminative features of the surface EMG signal. Autoencoders (AE)/Variable Autoencoders (VAE) are then trained for each of the seven deep features of the target motion intention. In addition, irrelevant motion intentions are detected according to the value of their reconstruction error. The recall rate of this method for the detection of irrelevant motion intentions exceeds 99% and the accuracy rate exceeds 99%.At the same time, we replaced the TCN with the LSTM model and compared the performance of the two after adding irrelevant motion discrimination. We collected data on seven goals and three unrelated motor intentions from seven experimenters for testing and completed an online experimental validation. The motion recognition accuracy of all the experimenters can be maintained above 86%.