Lower-body posture estimation with a wireless smart insole
Wing-kin Tam, Alan Wang, Baitong Wang, Zhi Yang · 2019
Typical optical marker and camera based systems for motion capture suffer from several limitations. They are restricted to indoor environments, have difficulties tracking multiple people simultaneously and require expensive camera setups. In this work, we present a new method for lower-body posture estimation with a wireless smart insole using end-to-end training of a deep neural network. Our model is able to predict the movement of the entire lower body (including the hip, knee, ankle and toe) accurately in a wide range of activities. Inference only takes 1.62ms and hence can be used in real-time. The proposed method can potentially provide a very efficient and portable solution for applications like sports analysis, rehabilitation and virtual reality.