Related Experiment Videos
Myoelectric signal classification using neural networks.
M Ungureanu1, R Strungaru, V Lazarescu
1Politehnica University of Bucharest, Dept. of Applied Electronics, Iuliu Maniu 1-3, Romania.
Biomedizinische Technik. Biomedical Engineering
|January 5, 2002
Summary
A feed-forward neural network effectively diagnoses spastic paralysis by classifying myoelectric signals. This method uses a 4th-order autoregressive model for accurate electromyography analysis.
Area of Science:
- Biomedical Engineering
- Neuroscience
- Signal Processing
Background:
- Spastic paralysis diagnosis relies on analyzing myoelectric signals (EMG).
- EMG signals exhibit stochastic behavior, often modeled using autoregressive (AR) models.
- Accurate AR model parameterization is crucial for reliable EMG analysis.
Purpose of the Study:
- To develop and evaluate a feed-forward neural network for diagnosing spastic paralysis.
- To determine the optimal order of the AR model for classifying normal and spastic paralysis EMG signals.
- To utilize the Hopfield algorithm for efficient AR model parameter calculation.
Main Methods:
- Utilized a two-layer perceptron feed-forward neural network.
- Recorded surface electromyography (EMG) signals using a surface electrode pair at 10 kHz.
- Employed a 4th-order autoregressive (AR) model, with parameters calculated via the Hopfield algorithm.
Main Results:
- The neural network successfully classified normal EMG from spastic paralysis EMG.
- A 4th-order AR model proved sufficient for the classification task.
- The Hopfield algorithm efficiently computed the necessary AR model parameters.
Conclusions:
- Feed-forward neural networks offer a viable approach for spastic paralysis diagnosis.
- A 4th-order AR model provides an effective balance between complexity and accuracy for EMG analysis.
- This methodology enhances the potential for automated and accurate diagnosis of neuromuscular disorders.