Upload
others
View
11
Download
0
Embed Size (px)
Citation preview
International Review on Computers and Software
(IRECOS) Contents: Hand Motion Gesture for Human-Computer Interaction Using Support Vector Machine and Hidden Markov Model by Suharjito, Fitra B. Adinugraha
374
Fuzzy Expert System for Classifying Pests and Diseases of Paddy Using Bee Colony Algorithms by Yovita Tunardi, Suharjito
381
A Framework for Social Media and Text-Based Content Analysis for Event Management Purposes by Qusai Abuein, Mohammed Q. Shatnawi, Muneer Bani Yassein, Radwan Batiha
388
Optimized Implementation of H.264/AVC Motion Estimation on a Mixed Architecture Using SynDEx-Mix by Oussama Feki, Thierry Grandpierre, Nouri Masmoudi, Mohamed Akil
395
Integration Electronic Patients’ Records with Open Life Sciences Datasets Using Semantic Web Tools by Bassam Najeeb, Bassel Al Khatib
403
Enhancing Relay Selection Scheme for Connecting VANETs to Internet Over IEEE 802.11p in Congested and Fading Environment Scenarios by D. Abada, A. Massaq, A. Boulouz
410
Key Frames Based Video Authentication Using Fragile Watermarking and Singular Value Decomposition by Assma Azeroual, Karim Afdel
420
Localization of Mobile Stations from ONE Base Station in GSM Systems by Khalid G. Samarah
427
Word Extraction from Arabic Handwritten Documents Based on Statistical Measures by Ayman Al-Dmour, Raed Abu Zitar
436
A New Compression Scheme Based on Adaptive Vector Quantization and Singular Value Decomposition by Imene Soussi, Mohamed Ouslim
445
Real-Time Electroencephalography-Based Emotion Recognition System by Riyanarto Sarno, Muhammad Nadzeri Munawar, Brilian T. Nugraha
456
ISSN 1828-6003Vol. 11 N. 5
May 2016
REPRINT
International Review on Computers and Software (IRECOS)
Editor-in-Chief: Prof. Marios Angelides
Brunel University London Department of Electronic and Computer Engineering Uxbridge, Middlesex - UB8 3PH, U.K.
Co-Editor: Dr. Harry Agius Brunel University London Department of Electronic and Computer Engineering Uxbridge, Middlesex - UB8 3PH, U.K.
Editorial Board:
Mikio Aoyama (Japan) Pascal Lorenz (France) Francoise Balmas (France) Marlin H. Mickle (U.S.A.) Vijay Bhatkar (India) Ali Movaghar (Iran) Arndt Bode (Germany) Dimitris Nikolos (Greece) Rajkumar Buyya (Australia) Mohamed Ould-Khaoua (U.K.) Wojciech Cellary (Poland) Witold Pedrycz (Canada) Bernard Courtois (France) Dana Petcu (Romania) Andre Ponce de Carvalho (Brazil) Erich Schikuta (Austria) David Dagan Feng (Australia) Arun K. Somani (U.S.A.) Peng Gong (U.S.A.) Miroslav Švéda (Czech) Defa Hu (China) Daniel Thalmann (Switzerland) Michael N. Huhns (U.S.A.) Luis Javier García Villalba (Spain) Ismail Khalil (Austria) Brijesh Verma (Australia) Catalina M. Lladó (Spain) Lipo Wang (Singapore)
The International Review on Computers and Software (IRECOS) is a publication of the Praise Worthy Prize S.r.l.. The Review is published monthly, appearing on the last day of every month. Published and Printed in Italy by Praise Worthy Prize S.r.l., Naples, May 31, 2016. Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved. This journal and the individual contributions contained in it are protected under copyright by Praise Worthy Prize S.r.l. and the following terms and conditions apply to their use: Single photocopies of single articles may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale and all forms of document delivery. Permission may be sought directly from Praise Worthy Prize S.r.l. at the e-mail address: [email protected] Permission of the Publisher is required to store or use electronically any material contained in this journal, including any article or part of an article. Except as outlined above, no part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. E-mail address permission request: [email protected] Responsibility for the contents rests upon the authors and not upon the Praise Worthy Prize S.r.l.. Statement and opinions expressed in the articles and communications are those of the individual contributors and not the statements and opinions of Praise Worthy Prize S.r.l.. Praise Worthy Prize S.r.l. assumes no responsibility or liability for any damage or injury to persons or property arising out of the use of any materials, instructions, methods or ideas contained herein. Praise Worthy Prize S.r.l. expressly disclaims any implied warranties of merchantability or fitness for a particular purpose. If expert assistance is required, the service of a competent professional person should be sought.
REPRINT
International Review on Computers and Software (I.RE.CO.S.), Vol. 11, N. 5
ISSN 1828-6003 May 2016
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved DOI: 10.15866/irecos.v11i5.8641
374
Hand Motion Gesture for Human-Computer Interaction
Using Support Vector Machine and Hidden Markov Model
Suharjito, Fitra B. Adinugraha
Abstract – Hand gesture recognition for human-computer interaction has become very popular
in recent times. The main problem of this technology is how the system can recognize the presence
of a gesture in a streaming video. In this paper, we propose a model that can recognize hand
motion gesture in avideo stream using Support Vector Machine and Hidden-Markov model.
Support Vector Machine has the advantage of generalizing classification. On the other hand,
Hidden Markov Model is a statistical model that is capable of modeling Spatial-temporal time
series. This system is divided into two main processes. First, this system recognizes hand posture
using SMV (static gesture recognition) and generates sequence observation which is used for the
second process later. The second process is recognizing dynamic hand gesture with the sequence
observation from the static gesture. The implementation shows that static hand gesture recognition
achieves average accuracy at 91% using testing dataset. Meanwhile,dynamically isolated hand
gesture recognition gets average accuracy at 89% using testing dataset. We also have tested the
system with continuous dynamic gesture using video stream. The system can recognize the gesture
very well with accuracy of 83%. This achievement shows that the model can be used in human
computer-interaction with specific supports such as Vector Machine and Hidden Markov Model
parameter. Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved.
Keywords: Gesture Recognition, Hand Gesture Recognition, Human Computer Interaction
I. Introduction
In daily life, gesture becomes one of main elements of
non-verbal communication. We use gesture to convey
some meaningful information. Gesture recognition is the
process in which gestures made by a user is made known
to the system. On the other hand, human-computer
interaction using motion gesture provides something
more natural, innovative and user-friendly [1]. Gesture
recognition has a wide range of applications, such as
virtual reality, augmented reality, sign language
recognition, video games, and medical application [2].
Recognition using the hand can be done using the
glove-based technique, device accelerator-based [3] and
visual-based technique [2]. While glove-based and
accelerator device sensing are very effective tools for
capturing hand motion, they are very expensive,
unnatural, and difficult to calibrate and setup the
procedure. On the other hand, the vision-based technique
provides natural gesture. It is inexpensive but still has
many challenges to improve the accuracy and speed [2].
One of the problems to achieve accuracy is how we
can extract the right feature that represents the hand
shape. Hand gesture recognition can be classified into
two types, static and dynamic. The static gesture is a
configuration or pose of the hand represented by a single
image while the dynamic gesture is represented by more
than one image [2]. In this paper, we focus on problems
regarding both of them.
After we have acquired static hand gesture, we use the
result to create a feature for the dynamic hand gesture. In
the static hand gesture, we use Support Vector Machine
(SVM) to classify single hand pose image with another
pose. Then, Hidden Markov Model (HMM) is used to
model dynamic hand gesture from the sequence of pose
hand image.
II. Related Work
In the last decades, there has been some research that
performs hand gesture recognition. Some of them
employ device based recognition like glove-based
gesture recognition [4], This glove-based gesture
recognition is very efficient in recognizing gesture
because the glove contains embedded sensors that can
detect motion very well. However, this method is
uncomfortable for the hand because this glove must be
attached to the wire. The other method using device-
based is MEMS accelerator recognition [3]. In this
method, the author tries to recognize motion gesture
through a device that can detect acceleration of hand
motion. Similar tothe glove-based gesture recognition,
this method is also efficient. However, we cannot
recognize the shape of the hand.
SVM is apopular model used for classification and
regression. In gesture recognition, SVM is commonly
paired with another feature extraction or model like HU
REPRINT
Suharjito, Fitra B. Adinugraha
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved International Review on Computers and Software, Vol. 11, N. 5
375
moments feature extraction. It has been used to classify
hand shape for static hand gesture [5].This pairing
method has an advantage of extracting moment of the
hand efficiently, such as hand orientation, angle, center
moment. The other method that has been paired with
SVM is Fourier descriptor used by Gamal et al. [2].
This method has some advantages, including HU
moments with additional fast processing because Fourier
descriptor only uses a limited number of feature for each
image. SVM is also often used to be combined with
another classification model as proposed by Demidova.
He combines SVM and fuzzy clustering algorithms [6].
Then, AbAzziz, who combines Fast Artificial Immune
and SVM for maximum load margin improvement [7].
The other method like HMM has been used to develop
Arabic continuous hand gesture from color image
sequence by applying trajectory of a single hand [8].
Malgireddy et al. [9] develop dynamic hand gesture.
They suggest a model explicitly (gesture grammar) and
use it to learn a model for each gesture. This method
divides gesture to sub-gestures and then uses the sub-
gestures to recognize the start point and end point of
gesture. With this method, Malgireddy uses HMM that
has been modified to use multiclass model and compare
HCRF. Hsiang-Yueh [10] and Hui-Shyong Yeo [11]
suggestthe use of YCrCb color space and convexity
defect character point to extract fingertip position and
perform robust dynamic hand gesture recognition with
simple algorithms. SVM HMM is also commonly used to
be combined with other classification models, like
proposed by Regan and Srivatsa.
They present a hybrid Canny Edge Detection and
Weibull Probability Density Function based on
Hierarchical Dirichlet Process HMM [12]. In this paper,
we focus on combining the base on these effective
methods. First, we suggest combining these advantages
from the previous research, such as using sub-gesture
from Malgireddy [9] with little modification.
We only use a single image to detect static gesture and
then perform the static gesture usingSVM. Second, the
used method is performing extract feature using
convexity defect method from Hsiang-Yueh to get robust
finger detection. Then, we also use HMM to recognize
continuous dynamics hand motion gesture. This process
will be explained in the methodology.
III. Methodology
In this section, we will present the proposed model of
human-computer interaction using hand motion gesture
recognition. The system is divided into four processes.
There are preprocessing, feature extraction, hand static
gesture and hand dynamic gesture recognition. Fig. 2
shows the proposed model.
III.1. Preprocessing
The purpose of this process is to prepare images for
the next phase.
We segment image color using YCrCb color space
information first and select the appropriate threshold
algorithm with additional threshold configuration to be
calibrated in the current environment. The default
threshold in the Eq. (1) [11]. Later, we will remove the
noise using dilate and mean filtering to smooth the
images. This process left the image of binary color:
77 ≤ Cb ≤ 127 and 133 ≤ Cr ≤ 173
where Y Cb Cr,, [0,255]
(1)
(a)
(c)
(b)
(a) (b)
(a)
(c)
(b)
(c)
Figs. 1. Hand Segmentation,(a) original image, (b) filter color image
(c) smooth image
The next step iscalculating contour of the hand
forfurther use in feature extraction.
Capture
Video
Image Filtering(YCrCb, MEDIAN,
COUNTOUR)
Feature Extraction(CONVEX HULL, CONVEXITY
DEFECTS, FINGERTIP)
Hand Classification(SVM)
Motion Classification(HMM)
Module Training
Motion Classification(HMM)
Module Training
Hand Classification(SVM)
GesturePrediction
(Left Click)
Models
Topology HMM
(A, B, )
Models SVM
(alpha, bias, SV, Kernel)
Class Label
(1-5-5-2-1-1-2)
Feature Vector1:50 2:32 3:40 ...
1:52 2:31 3:42 ...
Filterd ImageImage Frame
Fig. 2. Proposed model of hand motion gesture recognition
III.2. Feature Extraction Using Convexity Defect
Our proposed feature extraction method evolved from
previous research regarding hand tracking and gesture
recognition system [10], [11]. The result of preprocessing
is the contour of the hand. The contour is a list of the
point that represents a curve in an image.
REPRINT
Suharjito, Fitra B. Adinugraha
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved International Review on Computers and Software, Vol. 11, N. 5
376
First, we calculate polygon represented the
approximate to build contour [11]. Second, it is to extract
interior of the hand contour to gain more information
about hand shape image. This process can be done by
calculating palm center, and extracting convex hull and
convexity defect. Palm Center is an area determined as
the maximum inscribed circle inside the contour.
We calculate the maximum inscribed circle using
Voronoi diagram [13]. The maximum point of inscribed
circle lies on the point generated from the Voronoi
diagram. Next, the convexity defect will be illustrated in
Fig. 3. In this paper, we only use several convexity defect
points that match the following criteria:
- Depth point (Pd)and start point (Ps) must be above
the center of maximum inscribed circle (Ca)
- Each start point (Pa) and end point (Pe) must be
above Depth Point (Pd)
- Depth of each defect (ld) must be longer than palm
center radius (ra)
Fig. 3. Convex hull and convexity defect extracted from [11] hand
tracking and recognition system for human-computer interaction using
low-cost hardware
The next step is calculating or extracting the feature
from convexity defect that matches the criteria. In this
paper, we use two feature vectors.
The first feature is taken from the angle between depth
point (Pd) and end point (Pe) of each convexity defects
(θa), whilethe second one is taken from the depth
distance (distance between depth point and end point).
III.3. Static Gesture Recognition Using SVM
Support vector machine is a supervised classification
model that is often used to classify and regress. The basic
algorithm SVM is to separate the feature of training data
with a line or field separator, which is called hyperplane.
The function of the hyperplane is to maximize the
distance between the two pieces of data [2].
Basically, SVM is used to classify the two pieces of
data or classes. However, the application of SVM applied
is to classify more than two classes using method 1-vs-1
and 1-vs-all strategy 1-vs-1 by comparing one class with
another class so the class that has the highest value will
become the solution.
This method has andisadvantage. When the class that
will be classified multiplies, the time to classify will
increase, that can affect the performance of the
application. Moreover, the strategy of one-vs-all SVM
does createa model in which the model will detrain with
all class of existing dataset so that the decision will
depend on the model [14].The other ability of the SVM is
providing a transformation dataset from dimensional data
for other higher dimensional data defined by kernel
function [6].
Now, we have a set of feature hand shape images.
Then, the next step is to run the SVM classifier model
and get the best parameter to classify the images. SVM is
a supervised machine learning method used for
classification and regression.
The SVM algorithm separates the training data in
feature space by a hyperplane, which maximizes the
margin between two data. Hence, it is also known as
maximum margin classifiers [2]. SVM is based on the
principle of structural risk minimization (SRM). The
SRM induction principle has two main objectives. First,
it is to control the empirical risk on the training data, and
the second is to control the capacity of the decision
functions used to obtain this risk value [14].
There is an immediate problem arising from the
SVM’s original hyperplane formulation. It is not very
obvious how to make the model applicable to more than
two classes. Several approaches have been proposed to
overcome this limitation. Two of them are known as the
1-vs-1 and 1-vs-all strategies for multiple class
classification. For the decision problem over classes, 1-
vs-all requires the creation of classifiers.
Each is trained to distinguish one class from the
others. The decision is taken with a winner-takes-all
approach. However, there is no clear indication that this
approach results in an excellent decision. In the 1-vs-1
strategy, the problem is divided into c (c – 1) / 2 sub-
problems considering there are only two classes at a time.
At the decision phase, each machine casts a vote for
one of the classes and the label with the highest number
of votes wins.
This leaves the problem of evaluating an increased
number of machines every time a new instance is
classified – which can become troublesome or
prohibitive in time sensitive applications easily [15].
In this paper, we train and test four kernel types: RBF,
Linear, Polynomial, and sigmoid. Then, we propose ten
poses of static hand gesture as illustrated in Fig. 4. We
use multiclass classifier of SVM to classify ten classes of
static hand gesture. Fig. 4 shows static hand gesture
representing sequence observation label. This static
gesture is not used for human-computer interaction.
III.4. Dynamic Gesture Recognition using HMM
This step requires the recognition ofdynamic hand
gesture since the process considered is the dynamic
process and needs to be handled with the learning model
that can be accommodated with temporal time.
REPRINT
Suharjito, Fitra B. Adinugraha
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved International Review on Computers and Software, Vol. 11, N. 5
377
Fig. 4. 10 static hand gesture pose (later this pose used for sequence
observation label for dynamic hand gesture)
We propose HMM model [16] to accommodate this
dynamic motion gesture model. Hidden Markov Model is
one of the models based on the statistic. HMM is often
used in applications that intersect with time. HMM
possesses three main parameters, λ = (Π, A, B) [17],
where Π represents the vector beginning, A is the matrix
of transition while B is the matrix emission. Furthermore,
HMM has three major problems, which are Evaluation,
Decoding, and Training. However, these three problems
can be solved using the method of Forward-Backward
algorithm, Viterbi, and Baum-Welch [17].
In the application, HMM is often implemented using
three topologies. There are main Fully Connected
(Ergodic model) where all the states are connected as a
whole between the states; (Left-Right model) where each
state can only be returned to the state itself and in the
next state; (Parallel-Left-Right model) where each state
can only be repopulated themselves or towards the next
state. Shown in Figs. 6, the input of the model is the
result of sequence classification static hand gesture from
SVM. Figs. 5 show dynamic hand gestures in this
system.
Left Click
Right Click
Swipe
Right
Swipe
Left
Grap
(a)
(b)
(c)
(d)
(e)
Figs. 5. 5 Dynamic hand motion gesture that replacing mouse
functionality (random sample images from frame 1 until the end of
videos), (a) gesture for “Left Click”,(b) gesture for “Right Click”,(c)
gesture for “Swipe Left”,(d) gesture for “Swipe Right”
(e) gesture for “Grap”
(a)
(b)
(c)
Figs. 6. (a) ergodic, (b) left-right, (c) parallel-left-right [18] extracted
froma tutorial on hidden markov model
In this process, we train HMM model with HMM type
Ergodic and left-right then for each type of HMM.We
train it usingthe total hidden state from 2 until 9.
IV. Result and Discussion
Our proposed system is implemented in .net using
EmguCV wrapper library, libsvm, and accord hmm.
Fig. 7 shows the system implementation screenshot.
We use the standard webcam with 240 × 320 pixels
resolution. Moreover, we use conditional indoor lighting
environment. In this process, we use three scenarios to
test the system.
The first scenario that we find is the matching of SVM
parameter C and γ with the value from -3 to 3 for C
parameter and -15 to 3 for γ parameter.
We use four kernels (RBF, Linear, Polynomial, and
Sigmoid) in here.
Fig. 7. Hand motion gesture recognition system
REPRINT
Suharjito, Fitra B. Adinugraha
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved International Review on Computers and Software, Vol. 11, N. 5
378
The second scenario we see is the matching of HMM
parameter topology (Ergodic, left-right, and parallel-left-
right) and we test the number of the hidden state with the
value from 2-9. In this scenario, we will test it using
isolated hand motion gesture.
The third scenario is that we will test the best
parameter of SVM and HMM with continuous hand
motion gesture. Then, we calculate the accuracy with the
following equation:
(2)
We use two datasets that were created by our camera.
The first dataset contains 50 images used to train and 50
images to test each static gesture.
In this dataset, every image has various orientation
and scale. The second dataset contains 30 videos used to
train and 30 videos to test each isolated dynamic gesture
recognition.
On the other hand, we have two videos to test the
continuous dynamic hand gesture. One video has five
gestures that are performed sequentially, and another
video contains five gestures done randomly. In other
words, our dataset contains 1000 images and 302 videos.
Table I shows that kernel RBF with parameter C = 3 and
γ = -1 achieves maximum accuracy training at 100% and
93.4% for the testing dataset.
TABLE I
RESULT TEST SCENARIO 1 FOR SVM PARAMETER
Kernel Type C Value
(2x) γ Value (2x)
Training
Accuracy (%)
Testing
Accuracy (%)
RBF 3 -1 100 93.4
Linear 3 -15 99.4 89.4
Polynomial 0.5 1 76.6 79.4
Sigmoid 3 -3 97 83.8
In Table II topology parallel-left-right gets the
maximum accuracy from training HMM with the number
of hidden state = 4 and kernel SVM Polynomial.
On the other hand, Figs. 8, 9, 10, 11 show that the best
accuracy for hand motion gesture is HMM with topology
Parallel-Left-Right and with 2-4 number of hidden state.
The highest accuracy is achieved by using SVM
kernel polynomial with parameter C = 0.5 and γ = -1.
Then, Ergodic topology has the worst accuracy in all
SVM kernel.This is because the sequence observation
only needs various three sequences for each gesture,
whileErgodic topology, which has more transitions than
lef-right topology,was used to solve more complex
problem [17].
TABLE II
RESULT TEST SCENARIO 1USING RBF KERNEL SVM
HMM
Topology
Number of
Hidden State
Training
Accuracy (%)
Testing
Accuracy (%)
Ergodic 2 94 80
Left-Right 2 97 89.34
Parallel-Left-
Right 4 99 89.34
TABLE III
RESULT TEST SCENARIO 1 USING LINEAR KERNEL SVM
HMM
Topology
Number of
Hidden State
Training
Accuracy (%)
Testing
Accuracy (%)
Ergodic 2 51.34 38.67
Left-Right 2 94.67 83.34
Parallel-Left-Right 2 98 82
TABLE IV
RESULT TEST SCENARIO 1 USING POLYNOMIAL KERNEL SVM
HMM Topology Number of
Hidden State
Training
Accuracy (%)
Testing
Accuracy (%)
Ergodic 2 45.34 41.34
Left-Right 2 98 94
parallel-Left-Right 4 98 94.67
TABLE V
RESULT TEST SCENARIO 1 USING SIGMOID KERNEL SVM
HMM Topology Number of
Hidden State
Training
Accuracy (%)
Testing Accuracy
(%)
Ergodic 2 94 77.33333
Left-Right 2 96.66667 82.66667
Parallel-Left-Right 4 95.33333 82
Fig. 8 below shows the result of graph in finding the
best topology and number of hidden state in HMM.
The last scenario is to test the best model of SVM and
HMM for continuous dynamic hand gesture recognition.
The system shows a good result, that continuous hand
gesture with convergent dataset testing (videos that
contain only one gesture variance) can classify the
gesture with achievement up to 83%. Otherwise, testing
with videos that have random gesture achieves 63%. This
happens because convergent gesture videos have the
same sequence of image observation over all. Therefore,
HMM gets high probability to choose the same sequence
of image observation. If we test it with videos that have
the random sequence of gesture, the system gets false
recognition.
Fig. 8. HMM topology and number of hidden state testing result
for kernel RBF SVM
Fig. 9. HMM topology and number of hidden state testing result
for kernel LINEAR SVM
REPRINT
Suharjito, Fitra B. Adinugraha
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved International Review on Computers and Software, Vol. 11, N. 5
379
Fig. 10. HMM topology and number of hidden state testing result
for kernel POLYNOMIAL SVM
Fig. 11. HMM topology and number of hidden state testing result
for kernel SIGMOID SVM
V. Conclusion and Future Works
This paper suggests a system for human-computer
interaction using hand motion gesture. This system
achieves an average accuracy of about 97%. The highest
accuracy of static hand gesture is obtained from SVM
with kernel RBF, C = 3 and γ = -1.
On the other hand, the highest accuracy for dynamic
hand gesture is from HMM with topology parallel-left-
right using kernel polynomial with parameter C = 0.5 and
γ = -1. It reaches 98% of accuracy.
By using HMM topology parallel-left-right with four
hidden states, the system presents 97% of accuracy.
In the future, our system will focus on recognition
gesture spotting using a depth camera to create effective
skin color segmentation.
References
[1] M. Panwar and P. Singh Mehra, “Hand gesture recognition for
human computer interaction,” 2011 Int. Conf. Image Inf. Process.,
no. Iciip, pp. 1–7, 2011.
[2] H. M. Gamal, H. M. Abdul-Kader, and E. a. Sallam, “Hand
gesture recognition using fourier descriptors,” 2013 8th Int. Conf.
Comput. Eng. Syst., pp. 274–279, 2013.
[3] R. Xu, S. Zhou, and W. Li, “MEMS accelerometer based
nonspecific-user hand gesture recognition,” Sensors Journal,
IEEE, vol. 12, no. 5, pp. 1166–1173, 2012.
[4] P. Kumar, J. Verma, and S. Prasad, “Hand Data Glove: A
Wearable Real-Time Device for Human-Computer Interaction,”
Int. J. Adv. Sci. Technol., vol. 43, pp. 15–26, 2012.
[5] M. K. Bhuyan, D. Ajay Kumar, K. F. MacDorman, and Y.
Iwahori, “A novel set of features for continuous hand gesture
recognition,” J. Multimodal User Interfaces, 2014.
[6] Demidova, L., Sokolova, Y., Nikulchev, E., Use of Fuzzy
Clustering Algorithms Ensemble for SVM Classifier
Development, (2015) International Review on Modelling and
Simulations (IREMOS), 8 (4), pp. 446-457.
[7] Ab Aziz, N., Abdul Rahman, T., Zakaria, Z., Reactive Power
Planning for Maximum Load Margin Improvement Using Fast
Artificial Immune Support Vector Machine (FAISVM), (2014)
International Review of Automatic Control (IREACO), 7 (5), pp.
436-447.
[8] M. Elmezain, A. Al-Hamadi, J. Appenrodt, and B. Michaelis, “A
Hidden Markov Model-based continuous gesture recognition
system for hand motion trajectory,” 2008 19th Int. Conf. Pattern
Recognit., pp. 3–6, 2008.
[9] M. R. Malgireddy, J. J. Corso, S. Setlur, V. Govindaraju, and D.
Mandalapu, “A framework for hand gesture recognition and
spotting using sub-gesture modeling,” Proc. - Int. Conf. Pattern
Recognit., pp. 3780–3783, 2010.
[10] H. Y. Lai and H. J. Lai, “Real-Time Dynamic Hand Gesture
Recognition,” 2014 Int. Symp. Comput. Consum. Control, no. 1,
pp. 658–661, 2014.
[11] H. Yeo, B. Lee, and H. Lim, “Hand tracking and gesture
recognition system for human-computer interaction using low-
cost hardware,” Multimed Tools Applications 74(1) pp. 2687–
2715, 2015.
[12] Regan, D., Srivatsa, S., Mixed Pixel Wise Characterization Based
on HMM and Hyper spectral Image Gradient Enhancement for
Classification Using SVM-FSK, (2014) International Review on
Computers and Software (IRECOS), 9 (6), pp. 1017-1026.
[13] M. Schuster, “Largest Empty Circle Problem,”
https://www.cs.swarthmore.edu/~adanner/cs97/s08/papers/schuste
r.pdf pp. 28–37, 2008.
[14] L. Gericke, M. Wenzel, R. Gumienny, C. Willems, and C. Meinel,
“Handwriting recognition for a digital whiteboard collaboration
platform,” Proc. 2012 Int. Conf. Collab. Technol. Syst. CTS 2012,
pp. 226–233, 2012.
[15] C. R. Souza and E. B. Pizzolato, “Sign Language Recognition
with Support Vector Machines and Hidden Conditional Random
Fields: Going from Fingerspelling to Natural Articulated Words"
Machine Learning and Data Mining in Pattern Recognition,” vol.
7988, of the series Lecture Notes in Computer Science pp 84-98,
2013.
[16] A. Ramamoorthy, N. Vaswani, S. Chaudhury, and S. Banerjee,
“Recognition of dynamic hand gestures,” Pattern Recognition,
vol. 36, no. 9, pp. 2069–2081, 2003.
[17] M. Elmezain and A. Al-hamadi, “A Hidden Markov Model-Based
Isolated and Meaningful Hand Gesture Recognition,”Proceedings
Of World Academy Of Science, Engineering And Technology
(WASET) Vol. 31 July 2008 pp. 393–400, 2008.
[18] L. R. Rabiner, “A Tutorial on Hidden Markov Models and
Selected Application in Speech Recognition.” Proceeding of the
IEEE Vol. 77 No. 2 February 1989 pp. 257 - 286. 1989.
Authors’ information
Magister in Information Technology, Binus Graduate Program, Bina
Nusantara University, Jakarta, Indonesia.
E-mails: [email protected]
Fitra B. Adinugraha received his Bachelor
degree in Informatics majoring in Computer
Science from Institute Technology Telkom
(ITT), Indonesia in 2011. Presently, he is
pursuing the Graduate Program of Information
Technology at Bina Nusantara University,
Indonesia. His research interest includes
Human Computer Interaction (HCI), Natural
User Interface (NUI), Mobile Applications,
Cloud Storage, and Enterprise Architecture Design.
REPRINT
Suharjito, Fitra B. Adinugraha
Copyright © 2016 Praise Worthy Prize S.r.l. - All rights reserved International Review on Computers and Software, Vol. 11, N. 5
380
Suharjito is the Head of Information
Technology Department in Binus Graduate
Program of Binus University. He received under
graduated degree in mathematics from The
Faculty of Mathematics and Natural Science in
GadjahMada University, Yogyakarta, Indonesia
in 1994. He received master degree in
information technology engineering from
Sepuluh November Institute of Technology, Surabaya, Indonesia in
2000. He received the PhD degree in system engineering from the
Bogor Agricultural University (IPB), Bogor, Indonesia in 2011.His
research interests are intelligent system, Fuzzy system, image
processing and software engineering.
REPRINT
International Review on Computers and Software (IRECOS)
Aims and scope
The International Review on Computers and Software (IRECOS) is a peer-reviewed journal that publishes original papers on all branches of the academic Computer Science and Engineering communities. Thematic areas include, but are not limited to: Computer Science Theory, Methods and Tools Software engineering, algorithms and complexity, computational logic, formal methods, heuristics, mathematics and models of computation, programming languages and semantics. Computer and Communications Networks and Systems Network and distributed architectures and protocols, traffic engineering, resource management and Quality of Service, network monitoring and traffic measurements, wireless networks, personal and body area networks, vehicular networks, content and service-centric networking, multimedia communications and standards, energy efficient/green networks, opportunistic and cognitive networks. Computational Intelligence, Machine Learning and Data Analytics Human computer interaction, computational science, pattern recognition, computer vision, speech processing, machine intelligence and reasoning, web science, databases, information retrieval, visualisation, current applications domains, e.g. Healthcare and BioInformatics, and emerging application domains, e.g. big data. Security in Computer Systems and Networks Computer systems security, hardware and embedded systems security, security protocol design and analysis, cryptography and cryptanalysis, intrusion detection systems and techniques, user authentication techniques and systems. Hardware Design Computer architectures, parallel architectures, operating systems and signal processing.
Instructions for submitting a paper
The journal publishes invited tutorials or critical reviews; original scientific research papers (regular papers), letters to the Editor and research notes which should also be original presenting proposals for a new research, reporting on research in progress or discussing the latest scientific results in advanced fields; short communications and discussions, book reviews, reports from meetings and special issues describing research in any of the above thematic areas. All papers will be subjected to a fast editorial process. Any paper will be published within two months from the submitted date, if it has been accepted.
Papers must be correctly formatted, in order to be published. An Author guidelines template file can be found at the following web address:
www.praiseworthyprize.org/jsm/?journal=irecos Manuscripts should be sent via e-mail as attachment in .doc and .pdf formats to:
The regular paper page length limit is defined at 15 formatted Review pages, including illustrations, references and author(s) biographies. Pages 16 and above are charged 10 euros per page and payment is a prerequisite for publication. Abstracting and Indexing Information:
Cambridge Scientific Abstracts (CSA/CIG) Academic Search Complete (EBSCO Information Services) Elsevier Bibliographic Database - SCOPUS Index Copernicus - IC Journal Master List 2012: ICV 6.45
Autorizzazione del Tribunale di Napoli n. 59 del 30/06/2006
REPRINT
1828-6003(201605)11:5;1-7
REPRINT