The application of a neural network approach to solution of the EMG-torque relation in the ankle joint under isometric, supine conditions is addressed in this study. The backpropagation neural network model was used in this analysis to simulate a multilayer perceptron for solution to the muscle EMG-joint torque mapping. Measurements from 6 muscle sites are entered into the model as in the input, while the measured torque is entered into the model as the ideal output, to which the model Cutput is compared. The applied muscle signals may be considered the 'intent' of the system, while joint torque is the 'controlled' variable. It is expected that the results of this study will show feasibility for application of neural learning algorithms in the controller for intent recognition systems,such as robotic m anipulators, myoelectric prostheses, and functional electrical stimulators. I" The application of neural network models as an adaptive modelling technique is the primary focus of this study.The feasibility of designing a "neural network" which obviates the need to specify the EMG-torque relationship apriori wlll be addressed in this study. The electromyography (EMG) or muscle signal typlifies features of physiologic signals: real-time,non
[1]
Gordon D. Moskowitz,et al.
Myoelectric Pattern Recognition for Use in the Volitional Control of Above-Knee Prostheses
,
1981,
IEEE Transactions on Systems, Man, and Cybernetics.
[2]
A. Guez,et al.
A neuromorphic controller with a human teacher
,
1988,
IEEE 1988 International Conference on Neural Networks.
[3]
Neville Hogan,et al.
Myoelectric Signal Processing: Optimal Estimation Applied to Electromyography - Part II: Experimental Demonstration of Optimal Myoprocessor Performance
,
1980,
IEEE Transactions on Biomedical Engineering.
[4]
Daniel Graupe,et al.
Control of electrically-stimulated walking of paraplegics via above- and below-lesion EMG signature identification
,
1989
.