A calorie counter makes good sense as a smartwatch feature. Or it would, if smartwatches could track calories accurately. Sadly, for the most part, they do not. Now, researchers from Stanford University have developed new sensor technology that could improve the accuracy of wearable calorie tracking. The best part? The open-source instructions are free so that anyone can construct the affordable energy expenditure (EE) sensor system for themselves. 

In 2017, a study conducted by a separate Stanford University team found that popular smartwatches performed well at measuring heart rate, but failed to accurately measure EE. In that study, the lowest average device error rate was 27%, which hardly inspires confidence in calorie tracking. The highest average error rate was 93%, which indicates the calorie data from that device is flat-out useless.

Components of the new EE tracking system cost less than $100. The system includes two battery-controlled sensors placed on the upper leg and a microcontroller placed on the hip. A smartphone app could eventually replace the microcontroller. Moving the sensors from the wrist to the lower body boosts accuracy because leg movements account for more energy expenditure than the arms. 

Like many smartwatches, the system uses inertial measurement sensors. The research team designed special lightweight sensors made of inexpensive materials. The leg sensors could be sewn into smart pants or shorts to measure the acceleration and rotation of the leg as it moves.

The team tested the sensors using a group of participants representing a broad spectrum of ages, backgrounds, and fitness levels. The participants performed various activities requiring different levels of exertion. Compared to large-scale laboratory calorimetry equipment (which is the gold standard for measuring EE), the Stanford system demonstrated a 13% average error rate. While there’s still room for improvement, this is one-third the error rate of most smartwatches. The open-source technology offers data that’s accurate enough to rely on for many types of calorie tracking. 

A machine learning model that estimates calorie consumption based on sensor data from this cross range of participants will continue to improve accuracy. The researchers published their findings in Nature: Communications.

Better EE tracking could, of course, benefit users who wish to lose a few pounds. But precise wearable EE tracking has much broader implications. Personal tracking helps contextualize other biometric data that provides insight into more efficient workouts, individual metabolism, and many health conditions. And inexpensive and unobtrusive EE monitoring has clinical applications, such as tracking energy burn in patients with metabolic disorders or cardiac conditions.