# Muscle Strength Estimation of Key Muscle–Tendon Units During Human Motion Using ICA-Enhanced sEMG Signals and BP Neural Network Modeling

## TL;DR

This paper introduces a new method to accurately estimate muscle strength during human motion using enhanced sEMG signals and a neural network.

## Contribution

A novel approach combining ICA-enhanced sEMG signals and BP neural networks for high-accuracy muscle strength prediction.

## Key findings

- The proposed method achieved 98% localization accuracy with a sample size of 20.
- Muscle strength prediction reached 99% accuracy with a sample size of 100.
- The method reduced computational complexity while improving prediction efficiency.

## Abstract

Accurately predicting the muscle strength of key muscle–tendon units during human motion is vital for understanding movement mechanisms, optimizing exercise training, evaluating rehabilitation progress, and advancing prosthetic control technologies. Traditional prediction methods often suffer from low accuracy and high computational complexity. To address these challenges, this study employs independent component analysis (ICA) to predict the muscle strength of tendon units in primary moving parts of the human body. The proposed method had the highest accuracy in localization, at 98% when the sample size was 20. When the sample size was 100, the proposed method had the shortest localization time, with a localization time of 0.025 s. The accuracy of muscle strength prediction based on backpropagation neural network for key muscle–tendon units in human motion was the highest, with an accuracy of 99% when the sample size was 100. The method can effectively optimize the accuracy and efficiency of muscle strength prediction for key muscle–tendon units in human motion and reduce computational complexity.