USING MACHINE LEARNING TECHNIQUES AND WEARABLE INERTIAL MEASUREMENT UNITS TO PREDICT FRONT CRAWL ELBOW JOINT ANGLE: A PILOT STUDY

Authors

Angelo Macaro, The University of QueenslandFollow
Mark Connick, The University of QueenslandFollow
Emma Beckman, The University of QueenslandFollow
Sean Tweedy, The University of QueenslandFollow

Category

Methods

Document Type

Paper

Abstract

The purpose of this study was to evaluate the validity of using an artificial neural network (ANN) and inertial measurement units (IMUs) for estimating front crawl elbow angle in a laboratory environment. The study evaluates the validity of two models resulting from two swimmers who adopted different front crawl techniques. For each participant, data were collected from two IMUs placed on the arm during three minutes of ergometer-based swimming. These data were entered into an artificial neural network along with the target data which was elbow flexion angle from a camera-based motion capture system. The performance of each model was assessed by comparing the predicted elbow angle to the gold standard elbow angle during ten front crawl strokes collected separately from the training data. Root mean square difference (RMSD) between predicted and gold standard elbow angle across the ten stroke cycles was 7.75° for both participants. This pilot study demonstrates validity of using IMUs and artificial neural networks in a laboratory environment for estimating front crawl elbow angle in two swimmers who used different front crawl techniques.

Recommended Citation

Macaro, Angelo; Connick, Mark; Beckman, Emma; and Tweedy, Sean (2018) "USING MACHINE LEARNING TECHNIQUES AND WEARABLE INERTIAL MEASUREMENT UNITS TO PREDICT FRONT CRAWL ELBOW JOINT ANGLE: A PILOT STUDY," ISBS Proceedings Archive: Vol. 36: Iss. 1, Article 71.
Available at: https://commons.nmu.edu/isbs/vol36/iss1/71