PREDICTION OF JOINT KINETICS BASED ON JOINT KINEMATICS USING ARTIFICIAL NEURAL NETWORKS

Authors

Marion Mundt, RWTH Aachen UniversityFollow
Arnd Koeppe, RWTH Aachen UniversityFollow
Franz Bamer, RWTH Aachen UniversityFollow
Wolfgang Potthast, German Sport University, CologneFollow
Bernd Markert, RWTH Aachen UniversityFollow

Category

Computing/modelling

Document Type

Paper

Abstract

The high cost and low portability of measurement systems as well as time-consuming inverse dynamic calculations are a major limitation to motion analysis. Therefore, this study investigates predictions of joint kinetics based on kinematic data using an artificial neural network (ANN) approach. For this purpose, 3D lower limb joint angles and moments of twelve healthy subjects were calculated using inverse dynamics. Kinematic and anthropometric data was used as input parameter to train, validate and test a long short-term memory recurrent ANN to predict joint moments. The ANN predicts joint moments for subjects whose motion patterns are known to the ANN accurately. Although the prediction accuracy for unknown subjects was lower, this study proved the capability of ANNs to predict joint moments based on kinematic and anthropometric data.

Recommended Citation

Mundt, Marion; Koeppe, Arnd; Bamer, Franz; Potthast, Wolfgang; and Markert, Bernd (2018) "PREDICTION OF JOINT KINETICS BASED ON JOINT KINEMATICS USING ARTIFICIAL NEURAL NETWORKS," ISBS Proceedings Archive: Vol. 36: Iss. 1, Article 190.
Available at: https://commons.nmu.edu/isbs/vol36/iss1/190