Myoelectric Pattern Recognition And Control System Based On Surface Electromyography

There have been large quantities of amputees in China.Losing hand has brought expansive inconvenience to their life.The rebuild of grasping and manipulation functions is in urgent need.Conventional commercial prosthesis exploit multitude control which is unfortunately unnatural and leads to abandonment rate with peaks of 40%,This paper focus on developing a smart prosthesis controlled by EMG pattern recognition.Human intention can be decoded from the multi-channel EMG signals to control a multi-DOF hand.Following the typical control procedure,this paper designed the EMG acquisition devices,embeded EMG pattern recognition system and the hand driver system.First of all,multi-channel EMG acquisition module provides a man-machine interface.Pattern recognition decoder is responsible for the implementation of EMG algorithm whose embedded design method makes the system completely out of the computer.Finally,multi hand gestures and grasping modes are realized by the hand driver.Considering the control accuracy and This paper explores the embedded realization of two kinds of pattern recognition algorithms based on linear discriminant analysis and support vector machine.By introducing a certain data caching mechanism,the embedded program can realize onchip pattern recognition training and continuous EMG decoding.In addition,modular programming ideas can make the later development fucus on the realization of pattern recognition algorithm without worrying about EMG acquisition and preprocessing.In order to quantify the online recognition performance of the pattern recognition algorithm and verify the effectiveness of the control system,online pattern recognition experiments for 8 healthy subjects and one realtime control experiment for amputated patients were designed.Experiments show that when using the same classifier,TD features have better real-time control stability and response speed than AR features;Besides,the linear discriminant analysis and support vector machine algorithm can achieve similar good classification effect with average recognition rate up to 90%.Finally,equipped with the designed control system,the prosthetic arm can effectively assist the amputee to complete a variety of daily motions.