Development of a new low-cost computer vision system for human gait analysis: A case study.

Today, human gait analysis is commonly used for clinical diagnosis, rehabilitation and performance improvement in sports. However, although previous research works in the literature address the use of motion capture systems by means of optoelectronic sensors, Inertial Measurement Units (IMUs) and depth cameras, few of them discuss their conception, guidelines and algorithms for measuring and calculating gait metrics. Moreover, commercially available motion capture systems, although efficient, are cost restrictive for most of the low-income institutions. In this research work, a new computer vision-based system (CVS) for gait analysis is developed and proposed. The aim is to close the gap found in the literature about the design and development of such systems by providing the requirements, considerations, algorithms and methodologies used to develop a gait analysis system with acceptable precision and accuracy, and at low cost. For this purpose, a linear computer vision method based on the non-homogeneous solution of the calibration matrix was used. The spatio-temporal and angular gait parameters were implemented in the proposed system, and compared with those reported in the literature. The denoising of the spatial gait trajectories and the strategies to detect gait events, are also presented and discussed. The results have shown that the proposed system is satisfactory for human gait analysis in terms of precision, computational performance and low cost.