Distinguishing between Tip-Toe and Normal Gaits using Knee Angle Signal from the Skeleton Data Gathered by using OpenPose Module

Abstract

Background: This article offers a novel method for human gait recognition with OpenPose Module. The suggested technique emphasizes the detection of specific uncommon gait like tiptoe. Methods: For this purpose, an OpenPose module was employed to extract skeleton and joints, and the gaits in straight walking were obtained by using knee signal. Additionally, the similarity between the detected and normal (as reference) gaits was assessed with a dynamic time warping algorithm. Resultss: The algorithm outputs, normalizing factor, and unnormalized distance between input and reference signals were utilized to classify the normal and tiptoe gaits. Two features were selected as the most important features for classification. The proposed approach was tested among 70 individuals, which reached an accuracy of 92% between tiptoe and normal walking. Conclusions: The strong correlations with reference measurements support the recommended method for estimating tiptoe gait. The knee angle analysis during gait can be considered an identifier in other walking disorders. The mentioned difference may be a particular identifier for each walking disease. Accordingly, the present study results reflected the potential of the suggested approach to be used in identifying other walking disorders. Besides, walking evaluation can be done by getting similar to a typical gait.