1.1 Hand Gesture Recognition
Though the research has presented advancements in gesture recogni tion, further
advancements are required for making the gesture recognition process more
accurate. The existing research on gesture recognition can be classified into three
categories. The first category is the glove-based analysis. This employs sensors
attached to a glove that transduce finger flexions into electrical signals for deter-
mining the hand posture. This system requires the user to wear a cumbersome
device and carry a load of cables that connect the device to a computer. This hinders
the ease and naturalness with which a user can interact with a computer-controlled
environment. Potentially, the awkwardness in using gloves and other devices can be
overcome by video-based noncontact interaction techniques identifying gestures.
The second category comprises the models based on gesture images. Many
statistical, probabilistic, and other models have been proposed. One of the tech-
niques in this approach involves building a three-dimensional model of a hand. This
model is matched to the hand by one or more cameras and parameters corre-
sponding to the palm orientation and joint angles are estimated. These parameters
are used to perform corresponding gesture classification. Additionally, in order to
initialize the parameters the user places the hand in a specified position. The
parameters can also be manually instantiated.
The third category involves the analysis of drawing gestures. This method
usually involves stylus, which is used as an input device. This method also leads to
the recogniti on of written text. The majority of the work on HGR has employed
mechanical sensing, most often direct manipulation for the virtual environment and
occasionally for symbolic communication. Sensing the hand posture mechanically
involves a range of problems including reliability, accuracy, and electromagnetic
noise. Visual sensing has the potential to make gestural interaction more practical,
but potentially embodies some of the most difficult problems in machine vision.
This is because the hand is a nonrigid object and even worse self-occlusion is
commonly encountered.
1.1.1 Gesture Recognition Process
The live input of images can be taken using any normal web camera. As a video is a
sequence of images, the image sequence would be processed one by one. An image
frame may contain many undesired objects in addition to the hand gesture under
consideration. To identify the gesture correctly, the part of the image depicting the
hand needs to be separated from the part composed of the undesired objects. To
separate the skin from the image, a commonly used method involves applying a
suitable skin filter with an experimentally obtained threshold. This results in the
separation of the skin from the image while the undesired parts of the image can be
removed. In image processing, like other fields, the results are not extremely sharp
2 1 Introduction
我的内容管理
展开
