Identification and State Evaluation Based on Frontal Views of Walkers

Abstract

Walking, a basic human movement, is said to reflect an individual’s mental and physical state as well as individual characteristics. We discuss identifying walking gaits and evaluating mental and physical states using lower-limb trajectories extracted from frontal videos of walkers. We studied 3 mental and physical states - “normal,” “hurried,” and “tired” - using neural networks (NNs) were used for individual identification and state evaluation. From trajectories of the toes sampled from videos, we extracted static features (quantization) and dynamic features (displacement per unit time) and used these for teaching NNs. Experiments showed that trajectories showed features typical of individual walkers and their mental and physical states, with individual identification averaging 73% and evaluation 86.3%.