A wearable sensor module with a neural-network-based activity classification algorithm for daily energy expenditure...

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Machine Learning for Health

Background:

• Accurate estimation of energy expenditure (EE) is crucial for health monitoring and personalized fitness.
• Existing methods often lack precision in differentiating activities of similar intensity.
• Wearable sensors offer a promising avenue for continuous and objective EE assessment.

Purpose of the Study:

• To develop a wearable module and a neural network-based algorithm for precise energy expenditure estimation.
• To classify physical activities based on sensor data to improve EE regression models.
• To compare the performance of different neural networks for EE modeling.

Main Methods:

• Development of a wearable sensor module collecting acceleration and ECG signals.
• Implementation of a neural network-based algorithm including data preprocessing, activity classification, and feature selection.
• Utilizing sequential forward and backward search for efficient feature selection.
• Construction and comparison of energy expenditure regression (EER) models using Radial Basis Function Networks (RBFN) and Generalized Regression Neural Networks (GRNN).

Main Results:

• Validation of the wearable sensor module's effectiveness for EE estimation.
• Demonstration of the neural network-based activity classification algorithm's capability in enhancing EE estimation.
• Superior performance of the Generalized Regression Neural Network (GRNN) compared to the Radial Basis Function Network (RBFN) for EER modeling.

Conclusions:

• The proposed wearable sensor module and algorithm provide an effective solution for energy expenditure estimation.
• Neural network-based activity classification significantly optimizes EE estimation performance.
• GRNN emerges as a more effective model than RBFN for this specific application.