1
Rapid Hand Posture Recognition
Using Adaptive Histogram Template of Skin and Hand Edge Contour
Abstract: In this paper, we propose a real-time vision-based
hand posture recognition approach, based on appearance-
based features of hand. Our approach has three main steps:
hand segmentation, feature extraction and posture recognition.
For the hand segmentation, we introduce “Adaptive Histogram
Template of Skin” which tries to extract histogram of the
subject hand by sampling its color and texture. With this
template, we can use back projection method to find skin color
areas in an image. In the feature extraction step, we extract
global hand's features using hand's edge contour, and hand's
edge convex hull. The hand can be classified into one of the ten
posture classes in the recognition step. Each posture class has a
representative template which is used as reference for
comparing to subject hand features. This approach is simple
and fast enough to provide real-time recognition.
Keywords: hand posture recognition, skin detection, Hu
invariant moments.
1. Introduction
Hand posture recognition is a challenging computer
vision and pattern recognition issue [1,2]. The human
hand is a complex articulated object consisting of many
connected parts and joints. Considering the global hand
pose and each finger joint, human hand motion has
roughly 27 degree of freedom (DOF) [3].
Nowadays, the majority of human-computer interfaces
are mechanical devices such as keyboard and mouse and
joysticks. In recent years there has been a growing
interest in methods based on computational vision due to
its ability to recognise human gestures in a natural way
[4].
In the literature of gesture recognition, there are two
important definitions need to be cleared: hand posture
and hand gestures. A hand posture is defined solely by
the static hand configuration and hand location without
any movements involved. A hand gesture refers to a
sequence of hand postures connected by continuous
motions (global hand motion and local finger motion)
over a short time span [5].
Vision-based hand gesture recognition techniques can
be divided into two categories: appearance-based
approaches and 3D hand model-based approaches [6].
Appearance-based approaches use image features to
model the visual appearance of the hand and compare
these parameters with the extracted image features from
the input video. 3D hand model-based approaches rely on
a 3D kinematic hand model with considerable degrees of
freedom and try to estimate the hand parameters by
comparison between the input images and possible 2D
appearance projected by the 3D hand model [5].
1.2 Related Works
Hand gesture recognition has been addressed in [7]
and [8], but only for counting the number of stretched
fingers. Gesture recognition has been also addressed in
[9], but instrumented gloves are used, which simplifies
the recognition problem.
Moment invariants, and chain code histograms
methods has been approached in [10,11,12] respectively.
A 3D recognition approach requiring a multi-camera,
is proposed in [13]. A simpler 3D approach is addressed
in [14], which requires a 3D model with more than
twenty degrees-of-freedom, as well as a continuous,
error-free tracking of the hand (for updating 2D to 3D
correspondence).
2. Our Approach
In the present work, we focus on hand postures.
Furthermore, we use Appearance-based approach because
of its simplicity and advantage of real-time performance.
Our approach has three main steps: hand segmentation,
feature extraction and posture recognition.
For the hand segmentation, we introduce “Adaptive
Histogram Template of Skin” which tries to extract
histogram of the subject hand by sampling its color and
texture. With the adaptive histogram template of skin, we
can use back-projection method to find skin color areas in
an image.
In the feature extraction step, we extract global hand's
features using hand's edge contour and hand's edge
convex hull. From the hand’s contour and the hand’s
convex hull, we can calculate a sequence of contour
points between two consecutive convex hull vertices.
Ghassem Tofighi*, S.Amirhassan Monadjemi**, and Nasser Ghasem-Aghaee***
*University of Isfahan, tofighi@alum.sharif.edu
** University of Isfahan, monadjemi@eng.ui.ac.ir
*** University of Isfahan, aghaee@eng.ui.ac.ir
978-1-4244-9708-9/10/$26.00 ©2010 IEEE