Research On Multi-task Interactive Operation Of Prosthetic Hand Based On Electromyographic (EMG) Signal

Every year in the world,there are many amputation patients due to accidents,work-related injuries and other reasons.The amputation has a serious impact on their daily life and makes their daily activities very inconvenient;for upper arm amputees,they will be adapted after the amputation operation.Accepts cavity and prosthetic hands,but there are still many problems with prosthetic hands.For example,the widely used prosthetic hands have a single function and cannot meet the needs of patients to complete daily activities flexibly.And the prosthetic hands with more functions are expensive to fake,which leads to In daily life,patients basically only use normal hands and abandon prosthetic hands,and only use them as art hands.Therefore,the patient’s residual limbs are not used for a long time and cause muscle atrophy.In order to make the amputee more willing to use the residual limb in daily life to maintain the normal muscle level of the residual limb,in order to achieve the purpose of postoperative rehabilitation and make the daily life of the amputee more convenient,this paper proposes a method based on electromyographic signals and Deep learning real-time prosthetic hand control system.Electromyogram(EMG)is a weak electrical signal generated during muscle contraction.For the electromyogram data set,the wireless acquisition system of British Biometrics is used,equipped with 6 wireless EMG sensors,and the sensors are attached to the muscle to be measured.The location surface can accurately and effectively capture the signal change data of human muscles.The EMG signal data set established in this article contains three actions,namely,fist,two-finger pinch,and three-finger pinch.Each action corresponds to three different strength levels: mild,moderate,and severe.In the face of different daily activities,different strengths and gestures can be used to complete,improve the flexibility and comfort of the prosthetic hand,and make patients more willing to use the residual limb in daily life to achieve the purpose of rehabilitation.Secondly,for the recognition of EMG signals,the collected signals are time series signals,but for gesture recognition,it only takes a short time from the resting state to the completion of the gesture,but it will take a long time to maintain a stable gesture.,So it can be used as an image problem for recognition,so I chose to use CNN instead of LSTM,the recognition effect is better,and the training and recognition speed is faster.Since this study extracts gestures and strength categories from EMG signals and contains two outputs,multi-task learning is used instead of multi-classifier learning,training time is reduced by 1/3,only one model needs to be trained,and in this case The overall accuracy rate is slightly improved,and the effect is better.In the experiment of studying amputees,we put forward the idea of whether the cooperation of hands can enhance the perception ability of the residual limb,and conducted two sets of experiments.The control group experiment was the amputee using only the residual limb for data collection experiment,and the experimental group was amputation.The participants used both hands to conduct data collection experiments,combined with individual differences analysis of the experimental results,and concluded that the cooperation of hands can enhance the perception of residual limbs,and long-term experiments can be carried out to observe the rehabilitation effect.Finally,the entire online system is designed so that the amputee can control the prosthetic hand in real time through the deep learning trained model to achieve different strength grasping actions to complete daily activities,and test it on healthy people and amputees.