Estimating Internal Power in Walking and Running with a Smart Sock

he aim of this study is to investigate whether it is possible to estimate internal power in walking and running with a smart sock which is equipped with textile pressure sensors. Since commercially available smart socks are already used by runners to classify injury-prone running styles, such as running with low cadence and heel-striking, incorporating power measurement into the socks would make the usage of a separate power meter obsolete. While walking and running with different velocities and gradients on a treadmill, four subjects wore a pair of smart socks as well as a Stryd power meter as a reference system. The measurements from the pressure sensors were used to train regression algorithms, such as linear regression, trees of linear regressions (M5P), random forest, and k-nearest neighbors (KNN) to predict power. Preliminary results after a total of 42 runs show that depending on the actually used regression algorithm correlation coefficients between 0.75 and 0.99 and a mean absolute error between 1.5 and 21.8 Watts could be achieved. Although these results appear promising, the number of participants and test runs must be increased significantly in order to arrive at valid conclusions.