Estimation and Modeling of Forces, Power and Pressure Exerted on Feet During Gait Analysis Using Artificial Neural Network

Abstract

This study has been conducted to predict the forces and power excreted on the feet during Gait analysis. The weight of the human, the height of the human and weight of the shoe were the input process parameters for this study. 10 subjects were observed with two types of shoes namely heavyweight shoes and lightweight shoes. The data for the gait analysis was obtained by Kistler Force Plate System. In order to predict the accuracy of the measurements, Artificial Neural Network (ANN) has been used to analyze in swing and stance phase. The robustness of the model has been verified through 10 cross-validation method (k-fold). It was observed that the coefficient of determination (R²) was 78.2% for heavyweight shoes and 98.28% for lightweight shoes in medial-lateral direction (Fx). Similarly, it was 82.8% for heavyweight shoes and 87.1% for lightweight shoes in the anterior-posterior direction (Fy). For vertical direction (Fz) the values were 83.9% for heavyweight shoes and 92.1% in lightweight shoes. It has been observed that R² value for power was 87.9% for heavyweight shoes and 96.6% for lightweight shoes. To predict pressure the Artificial Neural Network has been used to develop the model in the swing and stance phase, hence validating the predicted parameter values with experimental ones. 18 subjects were observed with four types of insoles having different materials with lightweight shoes. Analysis of variance (ANOVA) has been used for selecting the input parameters. The robustness of the model has been certified through a ten cross-validation method (K-fold). Four types of insoles have been used in this study. R² for the pressure were 89.3%, 86.9%, 85% and 83% in Insole1, 2, 3 and 4 respectively. The outcomes of this research were tested with further subjects and results were found to be in close proximity.