EMG controlled bionic robotic arm using artificial intelligence and machine learning

Abstract

Electromyography is a unique approach for recording and analyzing the electrical activity generated by muscles, and a Myo-electric controlled prosthetic limb is an outwardly controlled artificial limb which is controlled by the electrical signals instinctively produced by the muscle system itself. Artificial Intelligence and Machine learning is very powerful in every technological field along with biomedical field. The purpose of this work is to utilize the power of Machine learning and Deep learning for predicting and recognizing hand gestures for prosthetic hand from collecting data of muscle activities. Although this technology already exists in the technological world but those are very costly and not available in developing countries. So, designing a cost effective prosthetic hand with the maximize accuracy is the major focus and objective of this work. We have also used a data set recorded by MyoWare Muscle Sensor which represents uninterrupted readings from 8 sensors. We have used Deep learning with the data set for predicting the following gestures which are handopen, hand-close, spherical-grip, and fine-pinch. Then we used some algorithms of Machine Learning which are K-nearest Neighbor (KNN), Support Vector Machine (SVM), and also the combination of KNN and SVM both for feature classification on data recorded with the 8-electrode surface EMG (sEMG) MyoWare Muscle Sensor. Using the combination of SVM and KNN We have accomplished a real time test accuracy of 96.33 percent at classifying the four gestures of our prosthetic hand. This paper also represents 3D modeling of the robotic hand and its control system using Autodesk 3D’s Max software, EMG MyoWare Muscle Sensor, Machine Learning and Deep Learning.