Deep Learning-Based Optimal Smart Shoes Sensor Selection for Energy Expenditure and Heart Rate Estimation

Area of Science:

• Sports Science
• Biomedical Engineering
• Wearable Technology

Background:

• Wearable devices enhance quality of life, with shoes offering a non-intrusive option.
• Smart shoes integrate sensors to monitor physiological and biomechanical data during activities.

Purpose of the Study:

• To develop and evaluate a deep learning model for estimating energy expenditure (EE) and heart rate (HR) using data from smart shoes.
• To implement a channel-wise attention mechanism for optimal sensor selection in EE and HR estimation.

Main Methods:

• Utilized smart shoes equipped with triaxial acceleration, triaxial gyroscope, and four-point pressure sensors.
• Employed a deep learning architecture that bypasses the need for separate preprocessing.
• Applied a channel-wise attention mechanism to dynamically weigh sensor contributions.

Main Results:

• The model achieved high accuracy in EE estimation (RMSE: 1.05 ± 0.15, MAE: 0.83 ± 0.12, R2: 0.922 ± 0.005).
• The model demonstrated strong performance in HR estimation (RMSE: 7.87 ± 1.12, MAE: 6.21 ± 0.86, R2: 0.897 ± 0.017).
• The z-axis of accelerometer and gyroscope sensors were identified as the most effective sensors for both estimations.

Conclusions:

• The proposed deep learning model effectively estimates EE and HR using smart shoe sensor data.
• The channel-wise attention mechanism enhances model performance by selecting optimal sensors.
• This technology offers a promising approach for monitoring physical activity and physiological responses.