Surface electromyography signal (sEMG) is the bioelectric signal accompanied by muscle contraction. For masterslave manipulation scenario such as patients with prosthetic hands, their upper limb sEMG signals can be collected and corresponded to the patient' s gesture intention. Therefore, using a slave manipulator that integrated with the sEMG signal recognition module, the master side could control it to make gestures and meet their needs of daily life. In this paper, gesture recognition is carried out based on sEMG and deep learning, and the master-slave control of manipulator is realized. According to the results of training, the network model with the highest accuracy of gesture classification and recognition can be obtained. Then, combined with the integrated manipulator, the control signal of the manipulator corresponding to the gesture is sent to the control module of the manipulator. In the end, a prototype system is built and the master-slave control of the manipulator using the sEMG signal is realized.

Surface electromyography signal (sEMG) is the bioelectric signal accompanied by muscle contraction. In gesture recognition, sEMG is a non-invasive, efficient and fast recognition method. For patients with hand amputation, their upper limb EMG signals can be collected, and these EMG signals correspond to the patient’s hand movement intention. Therefore, by wearing the prosthetic hand integrated with the EMG signal recognition module, patients with hand amputation can also make gestures meet their needs of daily life. In this paper, gesture recognition is carried out based on sEMG and deep learning, and the master-slave control of manipulator is realized. At the same time, gesture recognition can also be applied to remote control. Controlling the end of the manipulator at a certain distance with a specific gesture can complete some tasks in complex and high-risk environments with higher efficiency. Based on Convolutional Neural Network (CNN) and Gate Recurrent Unit (GRU), this paper constructs three neural networks with different structures, including single CNN, single GRU and CNN-GRU, and then train the collected gesture data set. According to the results of test set, the input type with the highest accuracy of gesture classification and recognition can be obtained. Among the three neural networks, CNN-GRU has the highest accuracy on the test set, reaching 92%, so it is used as the selected gesture recognition network. Finally, combined with the integrated manipulator, the EMG signals collected in real time by the myo EMG signal acquisition armband are classified by the upper computer, and the results are obtained. Then the control signal of the manipulator corresponding to the gesture is sent to the Arduino control module of the manipulator, and the master-slave control of the manipulator using the EMG signal can be realized.

A task motion trajectory usually needs to be determined for the training process and mechanism design for rehabilitation patients since they are not capable of providing a normal motion. In this paper, a machine-learning-based approach of gait trajectory prediction for lower limb rehab patients is proposed to provide the basis for the design of simple 1-degree-of-freedom (DOF) rehab mechanisms. First, a large amount of gait trajectories from various healthy volunteers are collected along with their body parameters, and a normalization method is presented to trim/expand these trajectory samples to a standard length and timing while retaining their shape and velocity information. Then, these normalized gait samples are clustered and regressed into a limited number of representative trajectories with K-means algorithm, and the cluster index is recorded as the label for each trajectory. Next, a genetic-algorithm-optimized support vector machine method is adopted to train a classifier for the trajectories, obtaining the correspondence between body parameters and cluster labels of gait trajectories. As a result, once a group of body parameters are input into the classifier, it can predict a most suitable gait trajectory for the specific patient. It shows that the accuracy of trajectory prediction reaches 96% both on training set and test set which verifies the effectiveness of the method. In the end, a 1-DOF gait rehab mechanism design example is provided to illustrate the application of the proposed method. Taking the predicted result from the classifier as the task motion trajectory for the synthesis of mechanisms, a 1-DOF six-bar mechanism is designed and the patient-mechanism matching can be realized.

Compared with traditional surgery, femtosecond laser minimally invasive surgery has many uncomparable advantages and will have a significant impact on the medical industry in the future. In this paper, a simple scene for laser minimally invasive virtual surgery training is designed, in which the testers can practice repeatedly until the basic operational requirements are met. The haptic device adopts the Geomagic Touch from American 3D Systems Company. Eight testers using Geomagic Touch handle perform four basic actions (clamping, adjusting posture, pushing / pressing, moving tiny objects) in the left interface virtual environment. Each action was performed 10 times by per tester. During the process of human-computer interaction, the position, attitude, speed, button and other information of the handle are collected in real time, and the collected data is saved in the form of text. The collected data is multivariate time series data. Based on the characteristics of multivariate time series data, this paper proposes a design of an evaluation system based on LSTM model to classify the collected data and evaluate the standard of surgical action according to the output probability of action classification.