PREDICTING THE 3D GROUND REACTION FORCE OF RUNNING USING LSTM BASED ON KINEMATIC PARAMETERS

• Main Author: Alvin
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/57128
• Institution: Institut Teknologi Bandung

Running has been a very common sport for many years, but still carries a high risk of injury. This risk is commonly referred to as Running Related Injury (RRI) which can be caused by many factors, one of which is the type of footwear. To analyze the effect of footwear on RRI, a system to analyze the kinematic and kinetic parameters of the running required. The kinematic parameter analysis system has been developed by Aditya Putra Khariza using markerless optical motion capture for running on a treadmill. However, there is still no kinetic parameter analysis system that can be used for running on a treadmill, so in this study, a kinetic parameter analysis system will be developed starting from the prediction of 3D ground reaction force (GRF) using the LSTM method based on the kinematic parameters from the optical motion capture. In this study, the LSTM method and dataset of running biomechanics, by Fukuchi et al., were used to train models with the help of MATLAB and Google Colabs. The training results showed that the predictions made on the test dataset have a good performance with average RSME value 1.91E-02, 4.45E-02, and 7.5E-03 for GRFx, GRFy and GRFz respectively. This result was smaller than the result of other deep learning methods such as ANN. Besides, this method also gave good mean coefficient correlation which is 0.993, 0.999, and 0.985 for GRFx, GRFy, and GRFz respectively. Therefore, in further research, this model can be used to predict GRF for running movements without an instrumented treadmill to save costs.