Labeling of Human Motion by Constraint-Based Genetic Algorithm

Abstract

This paper presents a new method to label parts of human body automatically based on the joint probability density function (PDF). To adapt to different motion for different articulation, the probabilistic models of each triangle different number of mixture components with MML are adopted. To solve the computation load problem of genetic algorithm (GA), a constraint-based genetic algorithm (CBGA) is developed to obtain the best global labeling. Our algorithm is developed to report the performance with experiments from running, walking and dancing sequences.