A New Algorithm for Labeling of Human Motion

Abstract

In this paper, we present a novel approach for the labeling of human motion based on a probabilistic model of body features and Constraint-Based Genetic Algorithm (CBGA), which learns the set of conditional independence relations among the body features through a fitness function. The approach allows the user to add custom rules to produce valid candidate solutions to achieve more accurate results with constraint-based genetic operators. We also extend these results to learning the probabilistic structure of human body to improve the labeling results, the handling of missing body parts, and the integration of multi-frame information to improve the accuracy rates. Finally, we analyze the performance of our proposed approach and show that it outperforms most of the current state of the art methods on a set of motion captured walking, running and dancing sequences in terms of quality and robustness. node ordering as a candidate solution in the population, and for each ordering, the solution is passed to K2, a greedy search algorithm, to obtain a network. In this work, the ordering and the conditional independence relations are also learnt separately, and the computation cost is very huge in this GA- based method. To resolve the computational efficiency problem of GA, search space reduction using constraints and supervised learning have been developed. Garofalakis (5) proposed a constraint-based algorithm to specify the expected tree size and accuracy in the search process. In other words, constraints can be used as a trade-off mechanism between the model accuracy and computational efficiency in the tree- building or tree-pruning process(6). In this paper, we modify posterior probability to resolve the missing body parts using CBGA. We also integrate information from multiple frames to improve the accuracy and robustness of labeling.