Human Action Recognition with Expandable Graphical Models

Human Action Recognition with Expandable Graphical Models

Wanqing Li (University of Wollongong, Australia), Zhengyou Zhang (Microsoft Research, USA), Zicheng Liu (Microsoft Research, USA) and Philip Ogunbona (University of Wollongong, Australia)
Copyright: © 2010 |Pages: 26
DOI: 10.4018/978-1-60566-900-7.ch010


This chapter first presents a brief review of the recent development in human action recognition. In particular, the principle and shortcomings of the conventional Hidden Markov Model (HMM) and its variants are discussed. We then introduce an expandable graphical model that represents the dynamics of human actions using a weighted directed graph, referred to as action graph. Unlike the conventional HMM, the action graph is shared by all actions to be recognized with each action being encoded in one or multiple paths and, thus, can be effectively and efficiently trained from a small number of samples. Furthermore, the action graph is expandable to incorporate new actions without being retrained and compromised. To verify the performance of the proposed expandable graphic model, a system that learns and recognizes human actions from sequences of silhouettes is developed and promising results are obtained.
Chapter Preview

Recent Development

A rich palette of diverse ideas has been proposed during the past few years on the problem of recognition of human actions by employing different types of visual information. A good review can be found in (Gavrila, 1999; Hu, Tan, Wang, & Maybank, 2004; Moeslund, Hilton, & Kruger, 2006; Moeslund & Granum, 2001). This section presents a brief review of the recent development in action recognition.

Study of the kinematics of human motion suggests that a human action can be divided into a sequence of postures. The sequence is often repeated by the same subject at different times or different subjects with some variations. Methods proposed so far for action recognition differs in the way that the postures are described and the dynamics of the posture sequence is modeled. In general, they fall into two categories based on how they model the dynamics of the actions: implicit and explicit models.

Complete Chapter List

Search this Book: