Gait Estimation from Anatomical Foot Parameters Measured by a Foot Feature Measurement System using a Deep Neural Network Model

Abstract

An accurate and credible measurement of human gait is essential in multiple areas of medical science and rehabilitation. Yet, the methods currently available are not only arduous but also costly. Researchers who investigated the relationship between foot and gait parameters have found that the two parameters are closely interrelated and suggested that measuring foot characteristics can be an alternative to the strenuous quantification currently in use. This study aims to verify the potential of foot characteristics in predicting the actual gait temporo-spatial parameters and to develop a deep neural network (DNN) model that can estimate and quantify the gait temporo-spatial parameters from foot characteristics. The foot features in sitting, standing, and one-leg standing conditions of 42 subjects were used as the input data and gait temporo-spatial parameters at fast, normal, and slow speed were set as the output of the DNN regressor. With the prediction accuracy of 95% or higher, the feasibility of the developed model was verified. This study might be the first in attempting experimental verification of the foot features serving as predictors of individual gait. The DNN regressor will help researchers improve the data pool with less labor and expense when some limitations get properly overcome.