3D Reconstruction of Periodic Human Walking Trajectories Based on Single View

Periodicity of a moving body is one of important characteristics in activity monitoring. We present a method to estimate the trajectory of human gait in 3D space from a single camera by exploring the periodicity of the human movement. Geometric constraints are established to characterize the periodic motion, which are invariant under viewing geometry variations. According to our analysis, these geometric constraints reduce the overall computational complexity of evaluating periodicity. Meanwhile, with the help of these geometric constraints, we develop a novel method to reconstruct the 3D motion trajectory from a single camera. Experimental results demonstrate the accuracy and robustness of the algorithm.