Myoelectric signal analysis using neural networks

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Signal Processing

Background:

• Myoelectric signal (MES) analysis is crucial for understanding muscle activity.
• Existing algorithms like SLS have limitations in processing speed for MES data.
• Neural networks offer potential for advanced signal analysis.

Purpose of the Study:

• To evaluate the efficacy of a discrete Hopfield network for extracting time-series parameters from MES.
• To compare the performance of the Hopfield network with the SLS algorithm.
• To explore the use of neural networks for classifying different types of muscular contractions using MES.

Main Methods:

• Utilized a discrete Hopfield network for functional minimization to extract MES time-series parameters.
• Employed a two-layer perceptron trained with back-propagation.
• Defined a two-dimensional signal space using signal parameters and power for classification.

Main Results:

• The Hopfield network demonstrated a faster rate of parameter extraction from MES compared to the SLS algorithm.
• The back-propagation trained perceptron successfully classified MES signals corresponding to different muscular contractions.
• Neural networks proved effective in analyzing MES data.

Conclusions:

• Discrete Hopfield networks offer a more efficient method for MES parameter extraction.
• Neural networks, particularly perceptrons, are suitable for MES analysis and classification tasks.
• Further research into neural network applications for myoelectric signal processing is recommended.