Handgrip estimation based on total variation denoising filtering for control applications

In many biomechanical studies and control applications, such as ergonomics studies, control of upper limb prosthesis, and sports performance is required handgrip force estimation for both monitoring and control purposes. As it was proven in previous works, features extraction from the extensor carpi radialis longus (ecrl) sEMG had a linear relationship with the gripforce of the hand. However, most of the developed estimations have shown high variation, which are not quite suitable for control applications. Therefore we propose a methodology to estimate the grip force, which models the extrated features as the handgrip force signal with the presence gaussian noise. In order to estimate the force, these features are filtered with a regularized optimization problem based on total variation denoising (TVD). Furthermore, since TVD is not a trivial minimization problem it was used ADMM algorithm as a meant to implement the proposed methodology. The developed methodology yielded promising results (ρ > 0.94 NRMSE < 0.07) between 30% - 50% MVC.