Activity Recognition and Energy Expenditure Estimation Using Wearable Sensors

Wearable sensors have transformed the objective measurement of human movement by enabling continuous monitoring of physical behaviours and metabolic energy use in real-world settings. Tri-axial accelerometers and inertial measurement units (IMUs) capture minute changes in motion, which are processed through threshold-based algorithms or advanced machine learning models to distinguish activity types such as walking, running and sedentary postures. Concurrently, estimation of energy expenditure has evolved from laboratory-bound respirometry to portable sensor arrays that infer metabolic cost from kinematic signals. Recent efforts have focused on optimising sensor placement, reducing algorithmic bias across demographic groups and achieving real-time estimations with minimal hardware. The convergence of hardware miniaturisation, open-source software and large-scale data from free-living cohorts is driving new insights into population health, personalised exercise prescriptions and weight-management strategies.

Research from Nature Portfolio

Recent studies have demonstrated that lower-limb sensor placement markedly improves energy estimation accuracy. A novel wearable system employing IMUs on the thigh and shank achieved a cumulative error of just 13 % in real-time metabolic power estimation across walking, running, stair climbing and cycling, substantially outperforming conventional wrist-worn devices. Such configurations converge more rapidly than physiological signals and enable time-varying activity monitoring. Foundational work on population-scale activity classification combined balanced random forest models with hidden Markov frameworks to label sleep and diverse activity modes from wrist accelerometry with around 87 % accuracy. These models revealed seasonal and sex differences in behaviour patterns and laid the groundwork for activity-informed public health guidelines.

Activity Recognition and Energy Expenditure Estimation Using Wearable Sensors publication trend

The graph below shows the total number of articles in activity recognition and energy expenditure estimation using wearable sensors across all publications each year (not limited to Nature Index journals).

Technical terms

Accelerometer: A sensor that measures acceleration forces along one or more axes to infer movement intensity and orientation.

Inertial measurement unit (IMU): A device combining accelerometers, gyroscopes and sometimes magnetometers to capture multi-axis motion dynamics.

Activity recognition: Automated classification of physical actions (for example walking or sitting) based on sensor signals.

Energy expenditure estimation: Quantification of metabolic energy use, typically inferred from kinematic data and algorithmic models.

Machine learning classification: Computational methods that learn discriminative patterns in labelled training data to assign new observations to predefined activity categories.

References

1. Sensing leg movement enhances wearable monitoring of energy expenditure. Nature Communications (2021).
2. Statistical machine learning of sleep and physical activity phenotypes from sensor data in 96,220 UK Biobank participants. Scientific Reports (2018).
3. A “one-size-fits-most” walking recognition method for smartphones, smartwatches, and wearable accelerometers. npj Digital Medicine (2023).
4. Wearable Leg Movement Monitoring System for High-Precision Real-Time Metabolic Energy Estimation and Motion Recognition.. Research (2023).