Low-power metabolic equivalents estimation algorithm using adaptive acceleration sampling.

Area of Science:

• Wearable technology
• Biomedical engineering
• Human motion analysis

Background:

• Accurate estimation of energy expenditure, quantified as metabolic equivalents (METs), is crucial for health monitoring.
• Traditional MET estimation methods often require high sampling rates, leading to increased power consumption in wearable devices.
• Developing low-power algorithms for MET estimation is essential for long-term physiological monitoring.

Purpose of the Study:

• To propose and evaluate a novel low-power algorithm for estimating metabolic equivalents (METs).
• To assess the algorithm's performance using triaxial acceleration data at adaptively changeable sampling rates.
• To quantify the power consumption benefits of the proposed algorithm compared to fixed, higher sampling rates.

Main Methods:

• Development of a MET estimation algorithm utilizing four adaptive sampling rates (32, 16, 8, and 4 Hz).
• Implementation of a switching mechanism for sampling rates based on synthetic acceleration patterns.
• Validation of the algorithm by applying it to one-day triaxial acceleration measurements.

Main Results:

• The algorithm successfully estimated MET values from triaxial acceleration data.
• Achieved a low root mean squared error (RMSE) for the calculated METs.
• Demonstrated significant power savings: 41.5% of 32 Hz consumption and 75.4% of 16 Hz consumption.

Conclusions:

• The proposed adaptive sampling rate algorithm offers an effective low-power solution for MET estimation.
• This approach balances accuracy and energy efficiency for wearable health monitoring systems.
• Further research can explore real-world validation across diverse physical activities and populations.