Inverse estimation of multiple muscle activations based on linear logistic regression

Area of Science:

• Biomedical Engineering
• Data Science
• Rehabilitation Technology

Background:

• Estimating muscle activity from movement data is crucial for clinical applications.
• Artificial Neural Networks (ANN) offer high estimation but suffer from data scarcity.
• Linear Regression (LR) models have limited generalization performance.

Purpose of the Study:

• To develop an improved statistical model for muscle activity estimation.
• To enhance the generalization performance of muscle activity estimation methods.
• To address the limitations of existing Linear Regression and Artificial Neural Network models.

Main Methods:

• Proposed a novel muscle activity estimation method using a linear logistic regression model.
• Compared the proposed method against traditional Linear Regression and Artificial Neural Network models.
• Conducted verification experiments with 7 human participants across various tasks.

Main Results:

• The proposed linear logistic regression model demonstrated superior generalization performance.
• The new method outperformed both Linear Regression and Artificial Neural Networks in the experiments.
• Effectiveness was validated across diverse tasks and participant movements.

Conclusions:

• The proposed linear logistic regression model offers a more robust solution for muscle activity estimation.
• This method shows promise for improving clinical applications where data is limited.
• Enhanced generalization performance is key for reliable muscle activity monitoring.