Speaker:

Julio E. Reátegui Pinazo

Abstract:

In many biomechanical studies and control applications, such as ergonomic studies, control of upper limb prosthesis, and sports performance. Is required handgrip force estimation for both monitoring and control purposes.

In most previous works, the force estimation process it’s done by gathering a large amount of sEMG data of one or more channels, and make an estimation based on some type of average. However the results present high variability in the estimated signal. On the other hand, other approaches, like learning algorithms, had shown better results, however the learning processes can be very time consuming. Therefore we propose a methodology to estimate the grip force, which models extracted features from one sEMG channel, as the handgrip force signal, but with the presence of gaussian noise.

In order to obtain a reliable signal, these features are filtered with a regularized optimization problem based on total variation denoising (TVD). Furthermore, since TVD is not a trivial minimization problem, the present paper shows the implementation of the proposed methodology using ADMM.

The simulations of the developed methodology yielded promising results ( > 0:94 NRMSE < 0:07) between 30%