Prediction Of The Angle Of Movement Of The Human Elbow Joint Based On Surface Electromyography Signals

Surface electromyography(sEMG)signals,a kind of biological electrical signals,can be detected from humans’ skin surface.it can reflect the state of nerve and muscle activity.While sEMG has the advantages of being convenient,accurate and non-invasive,it can be used in a wide range of research and application,such as the fields of rehabilitation medicine,sports medicine and intelligent robot and so on.Nowadays,the research on the identification of human behavior intention based on surface electromyography signals mostly focuses on the field of pattern recognition of body discrete movement.In a single mode of body movement,the researches on the details of body movements,such as motion speed,motion range,and the rotation angle,is less.If just using this kind of pattern classification results to control prosthetic limbs,we cannot achieve the result that the prosthetic joint can be easy for control like human joints.With the continuous development of information technology,predicting the continuous physical movement status using sEMG signals have become a key point for bionic limb control technology where sEMG signals are utilized.Therefore,we choose the arm elbow joint movement angle as the breakthrough point.And we are aimed to study the forecast method based of elbow motion angle based on sEMG signals,mainly explore the feature extraction of sEMG signals and network prediction model for the elbow motion angle.In this paper,the main contents are as follows:(1)construction of signal acquisition system.This paper firstly analyzes the produce mechanism and characteristics of sEMG signals in order to determine the measurement method of sEMG signals and shoulder joint angle,selecting suitable acquisition equipment,structures,and preparing acquisition platform for shoulder joint angle of the multi-channel and sEMG signals where the sEMG sensors,surface electrodes,MEMS sensors.Xsens inertial tracking devices and the acquisition software for sEMG signals and angle signals are all included.Then we design experiment plan to realize the signal collection of sEMG and angle of the elbowSignal processing and characteristic analysis.Firstly we analyzed sEMG signals and the main noise frequency segments and design filter to make the original signal more flat.After that we do some work for characteristics analysis.We extracted some representative characteristics from the time domain and frequency domain in order to take both time and frequency spectrum characteristics of the signals into account and provide the basis for the prediction of elbow joint movement angle model.Elbow joint movement angle prediction.We use the selected sEMG signal characteristic value as the input for the BP neural network,and make the elbow angle signal as the output of the neural network,constructing BP artificial neural network prediction model.And the BP neural network of prediction model was improved through the genetic algorithm.The prediction effect was analyzed,according to the error of the practical angle and the prediction results.The experimental results show that the prediction of elbow joint movement angle based on BP neural network are close to the actual angle and the average error range is about 8 degrees.On the other hand,the prediction average error is decreased to 5 degrees based on the new BP neural network which has been optimized by genetic algorithm.It shows that the genetic algorithm is benefit for improving the BP neural network,reducing the error and improving the prediction precision effectively.The proposed prediction method of elbow joint movement angle in this paper can spread to the shoulder,knee and other joints of the body.And this method combined with the pattern classification methods of body’s discrete action can also be applied to the bionic limbs control technology,improving the flexibility of limb movements.