International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075, Volume-9 Issue-1, November 2019
1161
Published By:
Blue Eyes Intelligence Engineering
& Sciences Publication
Retrieval Number: A4489119119/2019©BEIESP
DOI: 10.35940/ijitee.A4489.119119
Hand Gesture Recognition using Convexity Defect
Caroline El Fiorenza, Sandeep Kumar Barik, Ankit Prajapati, Sagar Mahesh
Abstract: Gestures are the simplest way of conveying a
message, rather simpler than verbal means. It is the most
primitive way of conversation. Gestures can also be the easiest
and intuitive way of communicating with a computer, they can be
used to communicate or convey information to computers, robots,
smart appliances and many other pieces of machinery. It can
eliminate the use of mouse and keyboard to some extent. The
gestures cited are basically the variable positions as well as
orientations of the hand. They can be detected by a simple
webcam attached to the computer. The image is first changed
into its corresponding RGB values and then to HSV values for
better handling and feature recognition. The hand is segregated
from the background using feature extraction. Then the values
are matched in proximity of the coded values. Then the region of
interest is calculated using the concept of convexity and
background subtraction. The convex defect helps to define the
contour efficiently. This method is invariant for different
positions or direction of the gesture. It is able to detect the
number of fingers individually and efficiently.
Keywords: RGB, HSV, hand gesture, background
subtraction, detection, feature extraction, convexity defect.
I. INTRODUCTION
Gestures have been used since primitive times to
convey messages in the simplest of fashions. Simple
messages, if not sophisticated were communicated by the
variable positions and orientations of the hand and other
parts of the body. They can be originated from any bodily
movement. The use of gestures as a form of communicating
with the digital media such as the PC and other smart
devices drives as the motivating force to analyse and record
the movements. The system needs to be visually intelligent
to capture and understand the movement. This includes
multi-disciplinary studies [1]. The gesture recognition
system is rapidly developing and is the stepping stone for
the advancement in smart appliances as well as IOT enabled
devices. The Human Computer Interaction (HCI) [1], is
simple and more intuitive as compared to the conventional
keyboard and mouse-based input. Gestures are easy to form
and remember and easier to produce.
This system finds application in simple PCs, smart
appliances, PDAs, medical equipment, public accessibilities,
offices with digital media, entertainment purposes etc. The
proposed system can be dissociated into the following
modules; sample image capturing, image pre-processing,
feature analysis, parameters extraction, classification and
recognition [3].
Revised Manuscript Received on November 06, 2019.
* Correspondence Author
Ms. Caroline El Fiorenza, Assistant Professor (O.G.), Department of
Computer Science and Engineering, SRM Institute of Science and
Technology, Chennai
Mr. Ankit Prajapati, Department of Computer Science and
Engineering, SRM Institute of Science and Technology, Chennai
Mr. Sandeep Kumar Barik, Department of Computer Science and
Engineering, SRM Institute of Science and Technology, Chennai
Mr. Sagar Mahesh, Department of Computer Science and Engineering,
SRM Institute of Science and Technology, Chennai
The paper is organised as follows, Literature survey,
materials and methods, features extraction, results and
discussion and at last the conclusion.
II. LITERATURE SURVEY
In one of the papers [1] the main contribution is to
introduce a forward movement finding strategy that
performs the separation and detection of hand signals all the
time. A stochastic approach is suggested to design a non-
motion process using CRFs model without data planning.
This system could efficiently identify signals and non-
motion models in tandem with forward spotting program.
Often, consider the time delay problems between the
detection and identification undertakings. The biggest
advantage of this method is that it improves efficiency and
also allows us to deal on complex issues that are abstracted
from the consumer and operate in the background, rendering
it more efficient and easier as well.
In another paper [2] the target of this paper is to plan a
hand motion acknowledgment framework that works
continuously and perceived manipulative hand signals. The
motions that are utilized in this acknowledgment framework
have particular significance. Every single one of these
signals speaks to a specific activity. This paper worries to
plan a component extraction technique that is invariant
overturn, scaling, interpretation, and direction-dependent on
minute element extraction strategy with the goal that the
framework can perceive hand motions caught in various
point or direction or size. A major demerit would be that the
proposed method is susceptible to errors, especially in
shapes like square and circular.
Human hand following [3], it requires predefining a
gathering of signal accumulation, from which getting the
main motion descriptor. In view of the mapping from
picture highlight space to signal space, we can evaluate
legitimately motion. By and large, such picture highlights
incorporate dab, line, corner or surface zone and so forth.
The use of this strategy needs setting up the mapping
connection between picture qualities and human hand act.
Nonetheless, close by moving procedure, hand's expressive
sharp changes make the mapping unquestionably profoundly
non-direct. This technique doesn't necessities ascertaining
3D motions, rather, it requires learning and preparing in
ample datasets which portray any conceivable signal.
Advantage of this would be objects dominating the image
will be represented as orientation histograms.
[4] We predominantly present the pre-processing steps
of motion pictures before we feed them into the DCNN. To
a limited extent A, we acquaint a shading offset calculation
with expelling the obstruction brought about by shaky light
sources and different components. Part B presents the signal
zone extraction technique dependent on Gaussian blend
model. Part C presents reproduction procedure of the entire