A Monocular Vision-Based Markerless Gait Analysis and Rehabilitation Assessment System

Abstract

Traditional gait analysis systems often suffer from high costs and operational complexity. To address these limitations, this paper proposes a single-view markerless gait analysis method and system for rehabilitation assessment. The system captures sagittal-plane walking videos using a monocular camera and employs Sports2D to extract two-dimensional human keypoint time-series data. Through temporal processing of keypoint trajectories, gait cycle segmentation, Heel Strike (HS) and Toe-Off (TO) event detection, and the estimation of gait parameters, including step length, cadence, walking speed, and lower-limb joint angles, are achieved. In addition, a gait visualization interface is developed to provide synchronized displays of videos, skeletal models, parameter curves, and tabular data. Experimental results demonstrate that the proposed method achieves stable performance in gait event detection and joint kinematic analysis, showing good agreement with both manual annotations and three-dimensional motion capture measurements. Owing to its low cost, ease of deployment, and user-friendly operation, the proposed system provides a practical tool for rehabilitation-oriented gait assessment.