IMU-Based Energy Expenditure Estimation for Various Walking Conditions Using a Hybrid CNN-LSTM Model.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Human Motion Analysis

Background:

• Energy expenditure estimation is crucial for monitoring physical activity intensity in healthcare.
• Existing methods often lack evaluation across diverse walking speeds and inclines.
• Inertial Measurement Units (IMUs) are key wearable sensors for activity monitoring.

Purpose of the Study:

• To develop and evaluate a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model for estimating energy expenditure.
• To assess the model's performance using only IMU data under various walking conditions (speed, incline).
• To compare fixed-size sequential data input with stride-segmented data input for the model.

Main Methods:

• A hybrid CNN-LSTM model was developed using IMU data.
• Experiments included level/inclined walking and level running on a treadmill.
• Model performance was evaluated using fixed-size sequential data versus stride-segmented data, considering sensor location and input format.

Main Results:

• The CNN-LSTM model achieved optimal performance with a two-second IMU data sequence from the lower body.
• The fixed-size sequential data approach showed comparable performance to stride-segmented data methods.
• The proposed model demonstrated practicality by not requiring heel-strike detection.

Conclusions:

• A hybrid CNN-LSTM model effectively estimates energy expenditure from IMU data during varied walking conditions.
• The model using fixed-size sequential data is a practical alternative to stride-segmented methods.
• This approach enhances the utility of wearable sensors for physical activity monitoring in healthcare.