An intelligent system for surface EMG-based position tracking of human arm movements for the control of manipulators

The design of a natural and a synergistic interface is essential to improve the human performance in a telemanipulation or a Virtual Reality (VR) system. Bio-electric signals, such as surface electromyogram (EMG), are being researched as alternate interfacing strategies for human arm position tracking and direct bio-control. An understanding of human joint dynamics is required to design a bio-electric interface for arm position tracking. Due to the complex nature of EMG signal and its relation to joint dynamics, an intelligent system is required to predict movements using surface EMG.; The overall objective of this study was to design a surface EMG based interface to track arm movements about the elbow joint. An intelligent system, consisting of neural networks and fuzzy logic, was developed to predict the elbow joint angle. The interface was evaluated on a computer simulated model of a robot. Normal subjects were asked to perform flexion-extension movements at various angles and speeds, as well as pronate their arms. Surface EMG signals were measured from the biceps muscle during flexion and from pronator teres during pronation of the arm. The joint angle at the elbow was measured using a goniometer. A signal processing module was developed to analyze the surface EMG signals and extract time varying magnitude-based parameters. The neural network was trained to predict the elbow joint angle using the magnitude and slope of the processed EMG signal. The fuzzy logic system computed an adaptive gain that compensated for changes in the biceps EMG signal due to variation in the speed of flexion. The interface was evaluated on a computer simulated model of a robot. The actual joint angle measured by the goniometer was compared against the joint angle predicted by the intelligent system and against the angle reproduced by the robot model. The coefficient of correlation between the actual joint angle and the predicted joint angle as well as the reproduced joint angle was calculated.; The intelligent system predicted the joint angle with average RMS errors of 5-25%. The correlation coefficient between the actual and the predicted joint angle was 0.92 with the arm in a supine position, 0.75 with the arm in a semi-prone position, and less than 0.5 with the arm in a prone position. The system accurately predicted for various angles and speeds of flexion, but the accuracy of the prediction decreased with the arm rotated (pronation). The average delay in tracking due to computations associate with the signal processing and the intelligent system was 0.2s. The robot model also reproduced the joint angle with RMS errors of 5-25%. The correlation coefficient between the actual and the reproduced joint angle was 0.9 when the arm was in a supine position. The overall delay in tracking due to the intelligent system and the robot model was 0.5s.; The study has demonstrated a unique and novel approach to position tracking and bio-control of telemanipulators and VR environments using surface EMG. It also represents a significant advancement in human joint dynamics. A successful attempt has been made to predict elbow joint angle using surface EMG with the aid of an intelligent system. The interface has several important applications in medicine, and perhaps the most significant one is towards the rehabilitation of paraplegics for myoelectric control of robotic assist devices.