Neural Networks for Muscle Forces Prediction in Cycling

Abstract

This paper documents the research towards the development of a system based on Artificial Neural Networks to predict muscle force patterns of an athlete during cycling. Two independent inverse problems must be solved for the force estimation: evaluation of the kinematic model and evaluation of the forces distribution along the limb. By solving repeatedly the two inverse problems for different subjects and conditions, a training pattern for an Artificial Neural Network was created. Then, the trained network was validated against an independent validation set, and compared to evaluate agreement between the two alternative approaches using Bland-Altman method. The obtained neural network for the different test patterns yields a normalized error well below 1% and the Bland-Altman plot shows a considerable correlation between the two methods. The new approach proposed herein allows a direct and fast computation for the inverse dynamics of a cyclist, opening the possibility of integrating such algorithm in a real time environment such as an embedded application.

A Neural Network Embedded System for Real-time Estimation of Muscle Forces

“…The neural estimator implemented in the model was refined starting from the one implemented in [16], and is composed by two parts: a Multiple-Input-Single-Output (MISO) NN, to calculate the θ3 angle, and a Neural System of nine MISO NN, to assess the muscular forces (one for each force).…”

“…In particular in these works both the kinematics and the dynamics required the solution of an optimization problem: the kinematic section required, for the computation of the angles between the elements schematizing the leg, the solution of a transcendental implicit equation; the dynamic section required the solution of a sparse and undetermined linear system, with 3 equations and 9 unknowns, subject to boundaries, through a cost function minimization, in order to estimate the muscular forces time trends. The solution of these problems using a set of Neural Networks for solution-prediction gave optimal results [16] when implementing the required algorithms on a standard workstation (Intel Core i7 with 16Gb RAM running Matlab in Windows 7 64-bit environment).…”

“…In a recent study [16] a new optimization algorithm based on artificial Neural Networks (NN) was proposed in order to reduce the computational complexity of the deterministic one previously used [10], while maintaining the quality of estimation. In particular in these works both the kinematics and the dynamics required the solution of an optimization problem: the kinematic section required, for the computation of the angles between the elements schematizing the leg, the solution of a transcendental implicit equation; the dynamic section required the solution of a sparse and undetermined linear system, with 3 equations and 9 unknowns, subject to boundaries, through a cost function minimization, in order to estimate the muscular forces time trends.…”

An Investigation of Fatigue in Stationary Biking with Application to Iliotibial Band Syndrome Development

“…While these models have predictive strength, they are limited in their generalizability to other activities and to the wider population (Noakes 2000). Lower level hierarchical models and neural networks are also investigated to predict muscle dynamics, but would require higher computational cost that could limit extension to real-time applications (Cecchini et al 2014;Heidlauf et al 2016). Through this investigation, the model development approach has been reevaluated.…”

“…Current human fatigue or performance models draw from medical records, sleep performance, and comprehensive activity tracking (R. Bini, Hume, and Croft 2011;van den Bogert et al 2013;Cecchini et al 2014). While these models have predictive strength, they are limited in their generalizability to other activities and to the wider population (Noakes 2000).…”