L. Alvarez et al. (Eds.): CIARP 2012, LNCS 7441, pp. 308–315, 2012.
© Springer-Verlag Berlin Heidelberg 2012
Hierarchical Elastic Graph Matching
for Hand Gesture Recognition
Yu-Ting Li and Juan P. Wachs
*
Department of Industrial Engineering, Purdue University, West Lafayette IN, U.S.A
{yutingli,jpwachs}@purdue.edu
Abstract. This paper proposes a hierarchical scheme for elastic graph matching
hand posture recognition. The hierarchy is expressed in terms of weights as-
signed to visual features scattered over an elastic graph. The weights in graph’s
nodes are adapted according to their relative ability to enhance the recognition,
and determined using adaptive boosting. A dictionary representing the variabili-
ty of each gesture class is proposed, in the form of a collection of graphs (a
bunch graph). Positions of nodes in the bunch graph are created using three
techniques: manually, semi-automatic, and automatically. The recognition
results show that the hierarchical weighting on features has significant discri-
minative power compared to the classic method (uniform weighting).
Experimental results also show that the semi-automatically annotation method
provides efficient and accurate performance in terms of two performance meas-
ures; cost function and accuracy.
Keywords: Elastic bunch graph, Graph matching, Feature hierarchy, Hand ges-
ture recognition.
1 Introduction
With the growing development of smaller, cheaper and versatile sensors, human-
computer interaction (HCI) relies more on natural communication, as among humans,
and less in standard interfaces such as the mice or keyboard [1,2]. This is reflected by
the users’ subjective satisfaction, the extent of the expressiveness and the overall ex-
perience perceived by the users of such systems [3]. Gestures are found extensively as
the main channel used to interact with computers in sign language interpretation [2],
assistive technologies [4], and game control applications [5]. Recently gestures were
adopted in new areas where sterility is essential to the task completion (e.g. browsing
medical images in the operating room) [6]. Nevertheless, to make gesture recognition
technologies to gain popularity in the HCI common market, high recognition accuracy
with low false alarms must be achieved through new pattern recognition techniques.
Elastic graph matching (EGM) is a technique used for object recognition [7],
where an object is represented by a labeled graph. The graph is matched against the
target image by computing filter responses at each node in the graph, and minimizing
*
Corresponding author.