Ontology-Based Semantic Gait Interpretation

Abstract

Human gait (manner of walking) and activity analysis from video sources is currently attracting attention in the computer vision community. The best current gait recognition system has a 90% identification rate under reasonable conditions, while recognition rates significantly decrease with a change of clothing, shoes, surface, or pose. In this paper, we present a framework for semantic gait interpretation and recognition using a knowledgebased method. The proposed approach consists of three phases: i) knowledge acquisition, ii) learning, and iii) categorization. We can solve the current gait recognition problem by linking the ontology of a meaningful human concept and locomotion to numerical data.