KKTM Ledang Compilation Journals

Compilation of Journal and Research
Articles (First Edition)

KKTM LEDANG

First Edition

Copyright © 2019 KKTM Ledang

First Edition 2019

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted in any forms, means, electronic, mechanical, photocopy, recording or otherwise,
without the prior written permission of the publisher.

KOLEJ KEMAHIRAN TINGGI MARA LEDANG

Serom 4 & 5, Jalan Serom-Bukit Gambir,

84410 Tangkak,
Johor Darul Ta’zim.

Tel : +606-9756200
Fax : +606-9756203

Website : http://ledang.kktm.edu.my/v2/
Facebook : KKTM LEDANG
Instagram : kktm_ledang_official

Cover design and layout by Norsyarizan Bin Shahri

Published by :

Teaching and Learning Material Publication Committee
Kolej Kemahiran Tinggi MARA Ledang
Serom 4 & 5, Jalan Serom-Bukit Gambir,
84410 Tangkak,
Johor Darul Ta’zim.

ii

FOREWORD

from the Director of KKTM Ledang

I would like to thank the Teaching and Learning Material Publication Committee for organizing
this 1st release Compilation of conference papers. It is a great pleasure to welcome all educators
and staff of KKTM Ledang to participate in any publication.

The compilation of conference and journal papers consists of 10 papers from various field of
studies such as electrical engineering, electronics engineering, mechatronics engineering,
biomedical engineering and engineering education. It is a platform that will enable the practitioner
leaders of engineer, technologies, scientist and educator in sharing knowledge, an opportunity to
interact with one another continuing and emerging issues relating to respected field studies in
general.

KKTM Ledang and MARA in general has the mission to promote and enhance learning in any
fields of engineering and technology. I believe it is important to create the right conditions for the
educators and students to express their creativity, while at the same time being confronted to the
realities of the world of work. The outcomes of these hard work from various field of research and
innovation hopefully will be beneficial and meaningful in advancing society, and to create
innovative ways for nurturing the future generation of engineers and educational leaders.

I hope that this 1st release compilation of conference and journal papers will be the platform to
promote continual exchange of ideas in a cross-cultural environment to foster and enhance research
in engineering and technology in particular, and in higher education in general to build a better
world for tomorrow’s generation. I also hope it would be a rewarding and fruitful experience for
all authors who contributes to this publication. Last but not least, I wish you a successful and
productive educator.

Fazli Rizal Ismail

iii

Teaching and Learning Material Publication Committee Year 2019

Ts. Kamil Bin Pongot
Norsyarizan bin Shahri
Dr. Nazrul Hamizi bin Adnan
Siti Marina binti Asari
Zulkarnain bin Shaharudin

Hanim binti Yusof
Wan Mohd Shukri bin Wan Salleh
Nik Md Hafizul Hasmie bin Mohamed Suhami
Khairunnisa binti Md Abdul Razak
Raja Mohd Aizat bin Raja Izaham

Zahari bin Hasan

iv

List of Publications

Analysis of EMG-based muscles activity for stroke rehabilitation.
Adnan, N. H., Suhaimi, R., Aswad, A. R., Asyraf, F., Wan, K., Hazry, D. & Razlan,
Z. M. (2014, August). In 2014 2nd International Conference on Electronic Design
(ICED) (pp. 167-170). IEEE.

Analysis of SOM and PCA Classifier for Finger Grasping Activities By Using
Glovemap.
Adnan, N. H., Wan, K., Bakar, S. A., Desa, H., Razlan, Z. M., & Ali, M. H. (2015).
International Journal of Innovative Computing Information and Control, 11(1), 163-
172.

Analysis of Object Grasp Force Using PCA BMU Approach.
Adnan, N. H., & Mahzan, T. (2015). International Journal of Mechanical &
Mechatronics Engineering IJMME-IJENS Vol: 15, No: 06.

Initial Results on Magnetic Induction Tomography Hardware Measurement
Using Hall-Effect Sensor Application.
Jaafar, N. H., Yazid, N. A. H. M., Zakaria, Z., Jumaah, M. F., Mansor, M. S. B.,
& Rahim, R. A. (2010). In 2010 IEEE EMBS Conference on Biomedical
Engineering and Sciences (IECBES) (pp. 9-12). IEEE.

Development of Red Blood Cell Analysis System Using NI Vision Builder AI.
Lias, J., Musa, R., Tomari, R., & Zakaria, W. N. W. (2015). ARPN Journal of
Engineering and Applied Sciences, 10(19), 8692-8698.

Screws Placement Effect on Locking Compression Plate (LCP) For Tibial
Oblique Fracture Fixation.
Izaham, R. M. A. R., Kadir, M. R. A., & Muslim, D. A. J. (2010). In 2010 IEEE
EMBS Conference on Biomedical Engineering and Sciences (IECBES) (pp. 236-
241). IEEE.

Student Approach to Learning in Programming Courses among Industrial
Mechatronics Engineering Technology Students.
Shahri, N., Rahman, R. A., & Hussain, N. H. (2014). In 2014 International
Conference on Teaching and Learning in Computing and Engineering (pp. 100-105).
IEEE.

v

Development of Single Axes Inertia Measurement Sensor Using
Complementary Filter Angle Estimation for Self-Balancing Platform
N. M. Hafizul Hasmie1, S. A. Junoh1 and R. M. A. Raja Izaham. International
Innovation Technology Exhibition & Conferences 2017.
Detection of Forgery Signature Using NI Vision Builder AI
N. F. N. Mahmod1, N. S. F. Hassan1, U. S. Jamaluddin1 and J. Lias1
SLaNS - Smart Laboratory Notification System
Nik Firdaus Nik Mahmod, Amimah Mohammad Ayub and Muhammad
Muhyiddin Mohamad Ibrahim.

vi

QG ,QWHUQDWLRQDO &RQIHUHQFH RQ (OHFWURQLF 'HVLJQ ,&(' $XJXVW 3HQDQJ 0DOD\VLD

Analysis of EMG-based Muscles Activity for Stroke
Rehabilitation

Rashidah Suhaimi2, Aswad A.R2, Nazrul H. ADNAN2,3 Khairunizam WAN1,2, D.Hazry1 Shahriman AB2, Juliana A.
Fakhrul Asyraf2 Abu Bakar3, Zuradzman M. Razlan1

2Advanced Intelligent Computing and Sustainability 1Centre of Excellence for Unmanned Aerial Systems
Research Group, School of Mechatronic (COEUAS)

Universiti Malaysia Perlis, Kampus Pauh Putra, 02600 2Advanced Intelligent Computing and Sustainability
Arau, Perlis, MALAYSIA Research Group, School of Mechatronic

3Bahagian Sumber Manusia, Tingkat 17 & 18, IbuPejabat Universiti Malaysia Perlis, Kampus Pauh Putra, 02600
MARA, Jalan Raja Laut, 50609 Kuala Lumpur, MALAYSIA Arau, Perlis, MALAYSIA

[email protected] 4Department of Multimedia School of Multimedia Tech &
Communication College of Arts and Sciences, Universiti

Utara Malaysia, 06010 Sintok, Kedah, MALAYSIA

[email protected]

Abstract— This paper presents 18 fundamental movements The outline of this research paper consists of section 2 for
for the rehabilitation of the stroke patient. The objective of this related research, section 3 for methods which present the
research is to develop the movement sequences which are methodologies of the proposed works. Section 4 includes the
suitable for the rehabilitation process and is focused on experimental result and the conclusion of this research is
hemiparesis sufferers which are the most common among expressed in Section 5.
stroke patients. The muscle activities are analyzed using
electromyography (EMG). 12 electrodes are attached to the II. RELATED RESEARCH
right arm of the subject includes deltoid, bicep, tricep, flexor
and extensor. The experimental results proof that it is likely to Recently, researchers had been using EMG
produce movement sequence for stroke rehabilitation based on (electromyography) device to study the activity of skeletal
each muscle activity. muscle contraction for various purposes for example for
motion recognizing, hand rehabilitation [8], classification
Keywords: Fundamental arm movement; rehabilitation; etc. EMG (electromyography) is an instrument able to
stroke patient; electromyography (EMG); motion sequence; transmit or detect the electrical signal generated by
electrically or neurologically energized muscle cells. The
I. INTRODUCTION output signals can be analyzed to detect the muscle activity
level and medical abnormalities even to analyze the
Arm rehabilitation in an important process after stroke to biomechanics of human.
regain movement skill lost. The main goal of stroke
rehabilitation is to regain independence and improve life’s Some of researcher studies the surface electromyography
quality. Arm rehabilitation must begin as early as possible of the upper arm of human to verify the variability in
after a stroke attack, right after completing all other different muscle location and different age group. The age
priorities such as stabilize the patient medical condition. group consists of three groups which are adolescents,
vicenarian and tricenarian. The researcher analyzed the
Various approaches that have been used formerly on mean, standard deviation and coefficient of variations to
hand rehabilitation such as virtual reality rehabilitation [1,2], study the muscle activities which are useful for
robotic rehabilitation [3,4], electrical simulation and data rehabilitation concerns [9].
glove [5]. Sandeep Subramanian et al. proposed a virtual
reality rehabilitation system by providing feedback to the Other researcher performs EMG signal analysis on
system, the virtual elevator scenes are suitable for the specific functional movements which are shots and passes
movement distance of patient and related to human body during playing basketball. The subjects are basketball
segment [6]. Meanwhile Shih Ching Yeh et al. suggest using players with different level of experience in playing
rehabilitation items of upper limbs includes three exercise of basketball. Flexor and extensor muscle were taken into
shoulder and elbow, the difficulty of the virtual game are consideration for each activity. The research proved that
adjustable to allow patient using different gesture [7]. EMG activation signal depends on the type of activity done
by muscle [10].
In this research, fundamental of arm movement data are
recorded using EMG system and the movements is
performed by a healthy subject. The output raw signals are
analyzed using Matlab to study the features of the signal.

‹ ,(((

QG ,QWHUQDWLRQDO &RQIHUHQFH RQ (OHFWURQLF 'HVLJQ ,&(' $XJXVW 3HQDQJ 0DOD\VLD

III. METHODS IV. RESULT
A. Experimental setup
A. Flow chart

Electromyography TABLE I. FUNDAMENTAL MOVEMENTS FOR DATA COLLECTION

Attachment of electrodes on Deltoid, No. Body part Fundamental Movements
bicep, tricep, flexor and extensor 1 Shoulder Shoulder Elevation
2 Shoulder Depression
Hemiparesis fundamental 3 Elbow Flexion and Extension
movement 4 Wrist Abduction and Adduction
5 External Rotation and
EMG signal processing Internal Rotation
6 Horizontal Abduction and
Fig. 1. Flowchart of the propose work Horizontal Adduction
7 Flexion and Extension
B. Signal Rectification 8 Supination and Pronation
Signal rectification is the process of translating the raw 9 Extension
10 Flexion
EMG signal to a positive polarity frequency. Two types of 11 Radial Deviation
rectification include full wave rectification and half wave 12 Ulnar Deviation
rectification. Full wave rectification converts all negative
values into positive values, this method preserves the energy Table 1 shows the muscle activities used in the experiment
signal for further analysis. While half wave rectification as the rehabilitation movement. All the movements were
deletes the negative signal leaving only the positive signal acquired by using EMG system as shown in figure 2.
and still can be used for statistical analyses [11].

C. Low Pass Filter Fig. 2. EMG system

Filtering a raw signal is an important process since the The ADInstruments EMG system used to acquire each
signal consists of many source of noise such as movement movement signal consists of 16 channels but only four
artifact, electrodes, cable movement artifact, electrostatic channels were designed for EMG, the capturing of the EMG
and radio waves. Suitable selection of filter is necessary to signal was done by using LabChart. Figure 3 shows the
avoid losing important information from the signal. electrode placement for the fundamental movement
Butterworth low pass filter is an ideal filter and able to get experiment. For the shoulder movement, 8 electrodes were
the closer approximation of the wanted frequency with the placed on deltoid, biceps and triceps. While for elbow and
right values of filter elements. The cut off frequency [12] wrist movements, 8 electrodes were placed on biceps,
selection is also necessary to make sure we only eliminate triceps, flexor and extensor. Each parts of muscle need a
the unwanted signal. pair of electrodes and one ground which is located at
another hand.
The formula of the Butterworth frequency response [13]
is given by

ቚ௏௏೚೔ೠ೙೟ቚ ൌ ଵ (1)
ଵାሺ௙Ȁ௙೎ሻమ೙

Where ܸ௢௨௧is the output voltage, ܸ௜௡ is the input voltage,
fstand for frequency, ݂௖ is the cut off frequency and n is the
elements filter number. The equation also can be written as
the transfer function ȁ‫ܪ‬ሺ݆‫ݓ‬ሻȁ as given below

ȁ‫ܪ‬ሺ݆‫ݓ‬ሻȁ ൌ ଵ (2)
ඥଵାሺ௪Ȁ௪೎ሻమ

In this experiment, the EMG signal is sample at the rate
of 1000 Hz/s with five repetition of rehabilitation movement
[14]. 5th order Butterworth low pass digital filter is used to
remove noises in the signal which the cut off frequency of
10 Hz and 0.001 normalization value.

QG ,QWHUQDWLRQDO &RQIHUHQFH RQ (OHFWURQLF 'HVLJQ ,&(' $XJXVW 3HQDQJ 0DOD\VLD

10 Low Pass Filtered EMG Signal
8
amplitude (V) 6
4
2 5 10 time (s) 15 20 25
0 (a)
0
Low Pass Filtered EMG Signal
10
ab amplitude (V)
8
Fig. 3. (a) Electrode placement for shoulder movement (b) Electrode 5 10 time (s) 15 20 25
movement for elbow and wrist movement 6
(b)
The subject who performed the stroke rehabilitation 4
movement is a healthy man in 24 years old. Figure 4 show Low Pass Filtered EMG Signal
the processing steps of EMG signal for elbow flexion 2
involving the bicep. The signal was rectified to obtain the amplitude (V)
signal which flow in positive direction and then was low 0
pass filter to smooth the signal. It can be seen that the signal 0 5 10 time (s) 15 20 25
after low pass filter is smoother and less repetition of data.
Rectified and filtered signal can be use for further analysis. 10 (c)
In this experiment, the most suitable filter is 5th order
Butterworth low pass filter of 10 Hz cut off frequency. 8 Low Pass Filtered EMG Signal
B. Experimental Results
amplitude (V) 6
(a)
4 5 10 15 20 25 30 35
(b)
2 time (s)

0 (d)
0
Low Pass Filtered EMG Signal
10
amplitude (V) 9
8
7 5 10 15 20 25
6
5 time (s)
4
3 (e)
2
1
0
0

10
9
8
7
6
5
4
3
2
1
0
0

(c)

Fig. 4. (a) Original unfiltered EMG signal (b) Rectified EMG Signal (c)
Low pass filtered EMG signal

QG ,QWHUQDWLRQDO &RQIHUHQFH RQ (OHFWURQLF 'HVLJQ ,&(' $XJXVW 3HQDQJ 0DOD\VLD

Low Pass Filtered EMG Signal a Rehabilitation Robot, 31st Annual International Conference of the
IEEE EMBS Minneapolis, 2009, pp. 7135-7138
10 [4] Jochen Radmer, Sami Hussein and Jorg Kruger. (2010) Depth Data
Based Human Motion Capture for Assessment of Robot Assisted Gait
9 Rehabilitation, IEEE Medical Technology Conference on
individualized Healthcare
8 [5] Nazrul H. ADNAN, Khairunizam WAN, Shariman Ab and Juliana A.
Abu Bakar., Principal Component Analysis For The Classification Of
amplitude (V) 7 Fingers Movement Data Using DataGlove “GloveMAP”.,
International Journal of Computer Engineering & Technology
6 (IJCET), Vol. 4, No. 2, 79-93, 2013
[6] Sandeep Subramanian, Luiz A Knaut et al. (2007) Virtual reality
5
environments for post-stroke arm rehabilitation, Journal of
4
NeuroEngineering and Rehabilitation 2007,4:20 doi:10.1186/1743-
3
0003-4-20
2
[7] Shih-Ching Yeh, Si-Huei Lee et al. (2012) Virtual Reality for Post-
1
Stroke Shoulder-Arm Motor Rehabilitation: Training System &
0 0 5 10 tim1e5 (s) 20 25 30
Assessment Method, 2012 IEEE International Conference on E-
(f)
Health Networking, Applications and Services, Beijing, 2012, pp 190-
Fig. 5. (a) Elbow flexion and extension (b)elbow pronation and supination
(c) Shoulder Abduction and adduction (d) Shoulder external and internal 195
rotation (e) Shoulder flexion and extension (f) Shoulder horizontal
abduction and horizontal adduction [8] K. Y. Ang, Y. Y. Huang, K. H. Low (2009) Electromyography

Five times repetition for each rehabilitation movement Analysis for Pre-clinical Trials off hand Rehabilitation Tasks using
can be observed clearly from the signals pattern. The peak
voltage shows the energy level of muscle contraction. Figure Design of Experiments, IEEE ICMA Proceedings, Changchun, China,
5 show the signals of deltoid obtained from several
rehabilitation movements which are elbow flexion and 2009, pp 915-920
extension, elbow pronation and supination, shoulder
abduction and adduction, shoulder external rotation and [9] Ahamed, N. U., Sundaraj, K., Ahmad, R. B. et al. (2012). Variability
internal rotation,shoulder flexion and extension finally the in surface electromyography of right arm biceps brachii muscles
shoulder horizontal abduction and horizontal adduction. The between male adolescent, vicenarian and tricenarian with distinct
movements were performed in pair according to suitability. electrode placement. In Sustainable Utilization and Development in
Engineering and Technology (STUDENT), 2012 IEEE Conference
V. CONCLUSION on (pp. 24-28). IEEE.

This research paper presents 18 fundamental arm [10] Pakosz, P. (2011). EMG signal analysis of selected muscles during
movements for stroke rehabilitation which was acquired shots and passes in basketball. J Health Promotion Recreation, 1, 9-
using ADInstrument Electromyography. A healthy subject 14.
was chosen to perform the arm rehabilitation movements
provided and guided by experience stroke therapist. [11] Information at

The sampling frequency which is 1000 Hz is suitable for http://en.wikipedia.org/wiki/Electromyography#Rectification
arm movement activities, the low pass filter with 10 Hz cut
off frequency also able to remove noise and artifact in the [12] M.Yoshida and M. Terao (2003) Suitable Cutoff Frequency of Low-
signals. The experimental results are useful to design a most
suitable motion sequence for stroke patient. In the future, the pass Filter for Estimating Muscle Forceby Surface Electromyogram,
functional movement based virtual environment will be
develop for the stroke patient to perform the rehabilitation Proceedings of the 2Sh Annual lnternational Conference of the lEEE
activities without stroke therapist.
EMBS, Mexico, 2003, pp 1709-1711
ACKNOWLEDGMENT
[13] Information at http://www.radio-electronics.com/
Thanks to all members of Advanced Intelligent
Computing and Sustainability Research Group, COEUAS [14] Rash, G. S. (2003). Electromyography fundamentals. Retrieved
for the ideas and help given. Not to forget School Of February, 4.
Mechatronics Engineering, Universiti Malaysia Perlis
(UniMAP) for the equipments and facilities provided.This
research is financed by MOSTI Science fund (9005-00059)
awarded to Universiti Malaysia Perlis.

REFERENCES

[1] Damasceno, E. F., Cardoso, A., & Lamounier Jr, E. A. (2013, March).
An Middleware for Motion Capture Devices Applied to Virtual Rehab.
IEEE Virtual Reality, pp. 171-172

[2] Nguyen, K. D., Chen, I. M., Luo, Z., Yeo, S. H., & Duh, H. L. (2011)
A Wearable Sensing System for Tracking and Monitoring of
Functional Arm Movement, IEEE/ASME Transactions On
Mechatronics, 2011, Vol. 16, No. 2, pp.213-220

[3] Nanda, P., Smith, A., Gebregiorgis, A., & Brown, E. E. (2009) Design
and Development of an Upper Extremity Motion Capture System for

International Journal of Innovative ICIC International ⃝c 2015 ISSN 1349-4198
Computing, Information and Control
pp. 1–14-03059
Volume 11, Number 1, February 2015

ANALYSIS OF SOM AND PCA CLASSIFIER FOR FINGER
GRASPING ACTIVITIES BY USING GLOVEMAP

Nazrul Hamizi Adnan1,2, Khairunizam Wan1,3, Shahriman Abu Bakar1
Hazry Desa3, Zuradzman Mohamad Razlan3 and Muhammad Hazwan Ali1

1Advanced Intelligent Computing and Sustainability Research Group
School of Mechatronic Engineering

3Centre of Excellence for Unmanned Aerial Systems (COEUAS)
Universiti Malaysia Perlis

Kampus Pauh Putra, 02600 Arau, Perlis, Malaysia
[email protected]; { khairunizam; shahriman; hazry; zuradzman }@unimap.edu.my

2Bahagian Sumber Manusia, Tingkat 17 & 18, IbuPejabat MARA
Jalan Raja Laut, 50609 Kuala Lumpur, Malaysia

Received March 2014; revised July 2014

Abstract. This research study presents the comparison of two classifier methods for
grasp recognition based on human grasping activities using selected objects (Bottle, Mouse
and Glue) without any limit in grasping style. Both classifier PCA and Self-Organizing
Maps (SOM) are employed to train the system by recognizing the objects using data glove
called as GloveMAP. The main purpose for this research is to differentiate the perfor-
mance between classifiers whiles the recognition task is done. At the end of the research,
the experimental results will show the difference in grasp recognition percentage between
PCA and SOM classifier to show the capabilities of classifiers in terms of suitability and
recognition performance.
Keywords: Data glove, Finger grasping, Grasping classification, Grasping recognition

1. Introduction. Recently, technological capability in recognizing the human activities
such as motion control [1,2], hand grasping [3,4] and robot grasping [5] is more and more
developed. They are demonstrated using such popular methods such as EMG [6], Data
glove [7-9], GloveMAP [10,11] and humanoid hand [12]. Nowadays, many researchers
classified the recognition of grasping motion into hand and finger joint motions. Napier
[13], Cutkosky [14] and Iberall [15,16] classified grasping into some clusters of categories
by considering the opposable thumb, and analyzed human hand motion especially for
robot hands based on the human grasping purpose.

Principal Component Analysis (PCA) generally functions as to reduce the dimension-
ality of dataset in which there are a large number of interrelated variables, while main-
taining them as much as possible in dataset changes. According to [17,18], PCA analysis
methods are capable to identify and express all dataset in such a way as to differentiate
their similarities and differences. Another classifier method such as Self Organizing Map
(SOM) motivated a new research area concerned with data interactions and it becomes
very popular among researchers. According to Kohonen [19], Self-organizing map (SOM)
classifier is an effective approach for high dimensional data analysis and processing. Both
classifiers (SOM and PCA) are capable to be the best classifier to finger grasping analysis
by using GloveMAP data glove (refer to Figure 1) in case to categorize the raw finger
movement data and are capable to supply an attractive alternative solution for recogniz-
ing data from the selected objects. The objective of this research is to verify the best

1

2 N. H. ADNAN, K. WAN, S. AB BAKAR ET AL.

Figure 1. Resistive interface glove (GloveMAP) [11,12]

recognition method for entire finger bending movement/motion signals using the selected
objects (Bottle, PC Mouse and Glue) that is recorded using GloveMAP. The improvement
to this research study is no limited to any subject grasping style or hand size. At the
same time, the result data might have difference between the subjects and also the shape
and size of an object grasped by every subject. Finally, the results are depending on the
GloveMAP data information received from finger object grasping before to be analyzed
using the proposed system.

This research paper is structured as follows. Section 2 addresses the literature review
including the approaches, applications and some problems regarding to process of recog-
nition task. Section 3 focuses on the methodologies and proposes some classifier study.
Meanwhile for the Section 4, it describes the results and discussions. Finally, Section 5
describes the conclusions and proposes some possible future works.

2. Literature Review. Manipulation of human fingers grasping makes possible the in-
teraction of human beings with the environment around them. The human finger hand, is
a complex and adaptable system, capable of both delicate and precise manipulation and
power grasping of any kind of objects [20,21]. According to Ratnasingam and McGinnity
[22] humans perform recognition tasks among to any object almost immediately and with
unlimited number of times. Human grasping of an object depends on feeling the contour
shape of an object by using fingers and the use of hand palm to grasp or move over the
object. The shape of an object can be described sufficiently using curvatures, angles and
surface contours [22,23].

According to Feix et al. [24], there are three type of grasping clusters compared with the
grasp taxonomy of human hand. They are regarded as grasp taxonomy from statistical
point of view (refer to Figure 2). One of the researches regarding to finger grasping is
about a prototype humanoid grasping developed by Dario et al. [25], which integrates
with major methods vision and tactile sensing for object manipulation using two fingers
and also a thumb. Meanwhile, according to Cobos et al. [26] the direct kinematics of
fingertips is used to grasp the objects and they also proposed the position and orientation
as the best methodologies for the study. Meanwhile, for the classifier for finger grasping,
Jerde et al. [27] stated PCA is the best classifier especially for the motionless position
synergy angle configuration of the physical posture and contour of human hand/fingers
whilst grasping the object. The authors also stated the use of PCA is capable to determine
postural synergies or kinematic movement of fingertips.

ANALYSIS OF SOM AND PCA CLASSIFIER 3

Figure 2. Grasp taxonomy proposed in [24]. In this table, all labels cor-
respond to power (label 1 to 7), intermediate (label 8 to 10), and precision
grasp (label 11 to 15) respectively.

However, according to [19], Kohonen stated the SOM classifier method is the best
among other classifiers in terms of the capability of producing the spatial organization
“internal representations” efficiently, and by using the SOM classifier various features of
input signals could be determined. SOM is one of the neural network methods that always
to be the famous method for computer applications analysis such as data extraction,
dimension reduction. SOM is known capable to provide the unsupervised learning network
which can map any entry mode to become one or two-dimensional discrete graphics of
the grasping features. The SOM is also capable to identify classes of grasping features
meaning after automatic clustering by the network, and SOM is necessary to simulate the
sample data.

3. Methodologies. A lot of work has been done in the field of principal component
and self-organizing [28,29]. Since these noticeable patterns should appear in the high-
dimensional joint space, a dimension reduction technique such as PCA could be effective.
By this research also it shows the SOM and PCA classifier could be as a fundamental
means for grasp analysis and synthesis based on the anatomy of human hand as shown
in Figure 3 below.

3.1. Self-Organizing Map (SOM). The advantage of SOM is to provide the unsuper-
vised learning network which can map any entry mode to become one or two-dimensional
discrete graphics of the grasping features. The SOM is capable to identify classes of grasp-
ing features meaning after automatic clustering by the network, and SOM is necessary to

Figure 3. Anatomy of the hand [30]

4 N. H. ADNAN, K. WAN, S. AB BAKAR ET AL.

simulate the sample data. A SOM does not need a target output to be specified unlike
many other types of network. Instead, there is a way how the node weights match the
input vector by training the weight vector. The process of SOM training occurs in several
steps:

(1) Weight initialization.
(2) Vector is randomly selected from the training data set and presented to the lattice.
(3) Every node is calculated and then to be examined in which one’s weights are most

similar to the input vector. At this stage the winning node is known as the Best
Matching Unit (BMU).
(4) Every single node in the lattice has their own neighbourhood, so at the same time
the BMU neighbourhood radius could be calculated. The value always starts large,
typically is set to the ‘radius’ of the lattice, but diminishes each time-step. Any nodes
found within this radius are deemed to be inside the BMU’s neighbourhood.
(5) Each neighbouring node’s (the nodes found in step 4) weights are adjusted to make
them more like the input vector. The closer a node is to the BMU, the more its
weights get altered.
(6) Repeat step 2 for N iterations.

The BMU training algorithm is based on competitive learning which is a particularly
same as the neural network supervised learning technique. To start the BMU features
learning, the first step is to initialize all the neurons weights in the dataset features
either to make the grouping values or sampled by the two largest principal component
eigenvectors of the training samples. In order to utilize the competitive learning training
technique, the sample dataset must be functioning as feeder to the features network by
calculating the distances between neurons to their positions with a distance function.
Euclidean distances between x and all the prototype vectors are computed, in order to
find the best matching neuron unit. The BMU is selected as the unit that is the nearest
to the input vector at an iteration t, using equation below:

∥x(t) − wc(t)∥ = mini ∥x(t) − wi(t)∥ (1)

Once the new BMU is generated then the winning neuron is identifying i∗ then the

“neighborhood” of the winning neuron could be calculated using the Kohonen rule [19].
Specifically, all such neurons i ∈ Θ(iq∗) are adjusted as follows:

Wi(q + 1) = Wi(q) + Θ(i, q)α(q)(p(q) − Wi(q)) (2)

where α(q) is a monotonically decreasing learning coefficient and p(q) is the input vector.
According to [24], it stated that the other method to simply determine the best matching
unit is using the node justification through all the nodes and the winning nodes could
be calculated using the Euclidean distance between each node’s weight vector and the
current input vector. The node with a weight vector closest to the input vector is tagged
as the BMU. V is known as the input vector and while W is called as the node’s weight
vector.

Dist = ∑i=n (3)
(V1 − W1)2

i=0

3.2. Principal Components Analysis (PCA). PCA was found useful in many ap-
plications such as data analysis, process monitoring and data rectification [29]. PCA is
a dimensionality reduction technique in terms of capturing the variance of the data and
it accounts for correlation among variables. The new axis coordinates are calculated by
converting the coordinate of the ordinary data. It is called as the space of Eigenfingers

ANALYSIS OF SOM AND PCA CLASSIFIER 5

(feature spaces). For example, let the dataset, consisting of p observation variables and
q observations for each variable be stacked into a matrix X ∈ Rq×p and it is expressed in

Equation (4):  x11 x12 · · · x1p 

X =  x21 x22 ··· x2p  (4)
... ... ... ...

xq1 xq2 · · · xqp

The principal component transform is defined by:

J = AT F (5)

A equals the Eigenfingers matrix after normalizing the covariance matrix of F . Then J

is called as diagonal covariance matrix of principal component shown in Equation (6):

 λ1 0 · · · 0 

Cj = ACX AT =  0 λ2 ··· ···  ; where CX = λiti; AT A = AT (6)
... ... ··· ...

0 · · · · · · λn

λ1 > λ2 > . . . > λn could be called as the eigenvalues of the covariance (some other
researchers call it as the diagonal covariance matrix) of F . The analysis of PCA that
could be used by both Eigenfingers and Eigenvalues are requisite. Whereas Eigenval-
ues can be simplified as Eigenvalues = Eigenfingers*original data. According to
[31], the analysis of the real numbers is dependent on both concepts (vectors and linear
transformations). Eigenfingers J of A and Eigenvalues λ can be determined as:

AJ = λJ (7)

and simplified as:

(A − λI)X = 0 (8)

The concept of Jacobi method [32] is applied where λ and A are calculated and I is

known as the identity matrix. Lastly, it is simple to find the Eigenfingers determinant as

shown in Equation (9).

det(A − λI) = 0 (9)

In particular, the activity of fingers grasping bending could reduce the number of fea-
tures needed for effective data representation by discarding the bending data. Equations
(10) to (12) show only the small variances and keep only those data terms that have a
large variances numbers [33]. For example, let λ1, . . ., λl denote the largest l eigenvalues
and associated Eigenfingers be denoted by Q1, Q2, . . ., Qx, respectively. The equation may
be written as:

∑I (10)
J¯ = AX QX (11)
(12)
X =1

1 ∑f xi
x¯ =
f i=1

δ2 = 1 ∑f − x¯)2
f (x1

i=1

For the calculation of dataset reduction the use of averages and standard deviations are
essential for data centering and reduction. x¯ is the arithmetic mean of each column, it

6 N. H. ADNAN, K. WAN, S. AB BAKAR ET AL.

is presented by Equation (11). The standard deviation is the square root of the variance
and it is presented by Equation (12).
3.3. Flow chart of works. Flow chart of works shown in Figure 4 provides overview of
the proposed system.

Figure 4. Flow chart of overall recognition system
3.4. Experiments. Six subjects (right-handed) participated in the experiment. Each
subject was fitted with a right-handed GloveMAP, which recorded all 3 flexible bend
sensors of the hand. Each subject participated in four experimental conditions. Figure 5
shows the activity involved in this research and all the subjects should follow the step to
extract the hand grasping data reading as follows:
a. Subjects were instructed to generate a set of hand grasping postures, designed to

reach all joint limits. Data from this condition was only used for calibrating the hand
grasping.
b. Subjects were asked to hold an object. The object was placed on a table and held
within 5-6 seconds and placed back to a table.
4. Results and Discussions. The experiment results from the study case analysis based
on object grasping by six subjects were presented. In this research study, the experiments
performed on three different objects (shown in Figure 6) with different object sizes. All
subjects under various grasping objects data were captured by MATLAB⃝R SIMULINK
software using GloveMAP. Meanwhile the process step of object grasping recognition

ANALYSIS OF SOM AND PCA CLASSIFIER 7

Figure 5. Object grasping activity

Figure 6. Selected objects (Bottle, Mouse and Glue)

Figure 7. The flow chart of the comparative studies between PCA and
SOM classifiers

was shown in Figure 7. The comparisons of different recognition algorithms between
two classifiers have always been a tough problem no matter what the classifier classified
or recognized. In order to comparatively evaluate the recognition performances of the
GloveMAP grasping activities, the classifiers were tested by using 20% of the total of 450
types of grasping style and another 80% were used to develop the database. The database
contains 360 grasping styles of 10 subjects. The confusion matrices for PCA and SOM

8 N. H. ADNAN, K. WAN, S. AB BAKAR ET AL.

Table 1. Confusion matrix for PCA algorithm on finger grasping

Object Bottle Mouse Glue PCA Recognition Rate (%)

Bottle 9 72 50.0

Mouse 1 15 2 83.3

Glue 6 39 50.0

Table 2. Confusion matrix for SOM algorithm on finger grasping

Object Bottle Mouse Glue SOM Recognition Rate (%)

Bottle 10 8 0 55.6

Mouse 3 14 1 77.8

Glue 5 1 12 66.7

Figure 8. Recognition percentage vs. object grasping

classifiers were shown on Tables 1 and 2. Table 1 shows that PCA recognized “Bottle”,
“Mouse” and “Glue” with the accuracies 50%, 83.3% and 50%, respectively. Meanwhile
Table 2 shows that SOM recognized “Bottle”, “Mouse” and “Glue” with the accuracies
55.6%, 77.8% and 66.7%, respectively.

In Figure 8, the X-axis represents the selected objects for this experiment (Bottle,
Mouse and Glue) and Y-axis shows the percentage of recognition for all classifiers. As
can be expected, the performance of three classifiers degrades with reduced amount of
training dataset.

5. Conclusion. The research paper proposed the comparative studies to classify finger
grasping by using low cost data glove called GloveMAP. From the analysis, the result
shows PCA and SOM are capable to classify finger grasping activities. Based on the
overall observations, it can be concluded that PCA has less recognition performance rate
compared with SOM. For the future works, the analysis could be extended to other objects
which are not limited to the size and shape of the object.

Acknowledgement. Special thanks go to all members of UniMAP Advanced Intelligent
Computing and Sustainability Research Group and School of Mechatronics Engineering,
Universiti Malaysia Perlis (UniMAP) for providing the research equipment and internal
foundations. This work was supported by the Majlis Amanah Rakyat (MARA) and

ANALYSIS OF SOM AND PCA CLASSIFIER 9

Science Fund Grant by the Ministry of Science, Technology and Innovation to Universiti
Malaysia Perlis (MOSTI 9005-00059).

REFERENCES

[1] I. L. Capisani and A. Ferrara, Trajectory planning and second-order sliding mode motion/interaction
control for robot manipulators in unknown environments, IEEE Transactions on Industrial Electron-
ics, vol.59, no.8, pp.3189-3198, 2012.

[2] S. Li, J. Yang, W. Chen and X. Chen, Generalized extended state observer based control for systems
with mismatched uncertainties, IEEE Transactions on Industrial Electronics, 2012.

[3] J. Kerr and B. Roth, Analysis of multifingered hands, Int. J. Robotics Res., vol.4, no.4, pp.3-17,
1986.

[4] D. Song, K. Huebner, V. kyrki and D. Kragic, Learning task constraints for robot grasping using
graphical models, IEEE Int. Conf. on Intelligent Robots and Systems, pp.1579-1585, 2010.

[5] A. Tsoli and O. C. Jenkins, Neighborhood denoising for learning high-dimensional grasping mani-
folds, IEEE Int. Conf. on Intelligent Robots and Systems, pp.3680-3685, 2008.

[6] C. Cipriani et al., On the shared control of an EMG-controlled prosthetic hand: Analysis of user-
prosthesis interaction, IEEE Transactions on Robotics, vol.24-1, pp.170-184, 2008.

[7] N. H. Adnan, K. Wan, S. Ab and J. A. A. Bakar, Principal component analysis for the classifica-
tion of fingers movement data using data glove “GloveMAP”, International Journal of Computer
Engineering and Technology (IJCET), vol.4, no.2, pp.79-93, 2013.

[8] C. Hager-Ross and M. H. Schieber, Quantifying the independence of human finger movements:
Comparisons of digits, hands and movements frequencies, The Journal of Neuroscience, vol.20,
pp.8542-8550, 2000.

[9] E. Todorov and Z. Ghahramani, Analysis of the synergies underlying complex hand manipulation,
Proc. of the 26th Annual International Conference of the IEEE EMBS, pp.4637-4640, 2004.

[10] T. Sugaiwa, G. Fujii, H. Iwata and S. Sugano, A methodology for setting grasping force for picking
up an object with unknown weight, friction, and stiffness, IEEE-RAS International Conference on
Humanoid Robots, pp.288-293, 2010.

[11] N. H. Adnan, K. Wan, S. Ab, J. A. A. Bakar and A. A. Aziz, PCA-based finger movement and
grasping classification using data glove “GloveMAP”, International Journal of Innovative Technology
and Exploring Engineering (IJITEE), vol.2, no.3, pp.66-71, 2013.

[12] N. H. Adnan, K. Wan, S. Ab and J. A. A. Bakar, Classification of finger grasping by using PCA based
on best matching unit (BMU) approach, International Journal of Advanced Research in Engineering
And Technology (IJARET), vol.4, no.2, pp.92-105, 2013.

[13] I. Napier, The prehensile movements of the human hand, Journal of Bone and Joint Surgery, vol.38,
no.4, pp.902-913, 1956.

[14] M. R. Cutkosky, On grasp choice, grasp models, and the design of hands for manufacturing tasks,
IEEE Transactions on Robotics and Automation, vol.5-3, pp.269-279, 1989.

[15] T. Iberall, Human prehension: Three dexterous hands in one, International Conference on Robotics
and Automation, vol.16, pp.258-299, 1987.

[16] T. Iberall, Human prehension and dexterious robot hand, International Journal of Robotics Research,
vol.16, no.3, pp.285-299, 1997.

[17] M. Turk and A. Pentland, Eigenfaces for recognition, Journal of Cognitive Neuroscience, vol.3,
pp.71-86, 1991.

[18] S. Cobos, M. Ferre, M. A. S´anchez-Ura´n, J. Ortego and R. Aracil, Human hand descriptions and
gesture recognition for object manipulation, Computer Methods in Biomechanics and Biomedical
Engineering, vol.13, no.3, pp.305-317, 2010.

[19] T. Kohonen, Self-organized formation of topologically correct feature maps, Biological Cybernetics,
vol.43, no.1, pp.59-69, 1982.

[20] I. Kapandji, The Physiology of the Joinrs, Churchill Livingstone, Edinburgh, 1970.
[21] E. Y. Chao, K.-N. An, W. P. Cooney and R. Linscheid, Biomechanics of the Hand: A Basic Research

Study, World Scientific Pub. Co., 1989.
[22] S. Ratnasingam and T. M. McGinnity, Object recognition based on tactile form perception, IEEE,

2011.
[23] D. Burke, S. C. Gandevia and G. Macefield, Responses to passive movement of receptors in joint,

skin and muscle of the human hand, J. Physiol., pp.347-361, 2007.

10 N. H. ADNAN, K. WAN, S. AB BAKAR ET AL.

[24] T. Feix, O. Bock, H. Gmbh, H. Schmiedmayer, J. Romero and D. Kragic, A comprehensive grasp
taxonomy, Robotics, Science and Systems Conference: Workshop on Understanding the Human Hand
for Advancing Robotic Manipulation, 2009.

[25] P. Dario, C. Laschi, M. C. Carrozza, E. Guglielmelli, G. Teti, B. Massa, M. Zecca, D. Taddeucci and
F. Leoni, An integrated approach for the design and development of a grasping and manipulation
system in humanoid robotics, Proc. of the 2000 IEEE/RSJ International Conference on Intelligent
Robots and Systems, vol.1, pp.1-7, 2000.

[26] S. Cobos, M. Ferre, M. A. S´anchez-Ur´an, J. Ortego and R. Aracil, Human hand descriptions and
gesture recognition for object manipulation, Computer Methods in Biomechanics and Biomedical
Engineering, vol.13, no.3, pp.305-317, 2010.

[27] T. E. Jerde, J. F. Soechting and M. Flanders, Biological constraints simplify the recognition of hand
shapes, IEEE Trans. Biomed. Eng., vol.50, pp.265-269, 2003.

[28] J. Park and J. Cheong, Analysis of collective behavior and grasp motion in human hand, International
Conference on Control, Automation and Systems, pp.2514-2518, 2010.

[29] J.-K. Lee, K.-H. Kim, T.-Y. Kim and W.-H. Choi, Non-linear principle component analysis using
local probability, Proc. of the 7th Korea-Russia International Symposium, pp.103-107, 2003.

[30] http://www.computer.org/csdl/trans/th/2013/01/tth2013010106-abs.html.
[31] http://en.wikipedia.org/wiki/Eigenvalues and Eigenvectors.
[32] http://en.wikipedia.org/wiki/Jacobi method.
[33] S. Haykin, Neural Networks and Learning Machine, Pearson, 2009.

International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:15 No:06 59

Analysis of Object Grasp Force Using PCA-
BMU Approach

Nazrul H. ADNAN, Mahzan T., Khairunizam WAN, and Nabilah H.E

 Dataglove is also known as parts of “Haptic Science”,
Abstract— This paper reports on the analysis of human which is give meaning as science of applying tactile
sensation to human interaction through computer.
studies for the purpose of finger grasp using Gaussian-
Principal Component Analysis (PCA) and PCA-Best Matching Datagloves are one of several types of electromechanical
Unit (BMU). The method purposes to find the best grasping devices used in haptics applications. Haptics refers to
feature data using GloveMAP which is based on the fingers sensing and manipulation through touch while haptic device
adapted grasping force movement. Gaussian filter method usually is a device which involves all aspect of information
functions to smoothen the force signal at the same time capable acquisition and object manipulation through humans,
to eliminate / remove the overshoot signal and suitable to be machines or combination of them. It is call physical contact
used for filtering grasping force input signals while minimizing between computer and user through a device that sensing
the rise and fall time of the grasping object. Meanwhile for the movement of body such as mouse, joystick, keyboard or an
finger grasping group features, the method of Best Matching input / output device. Here, researches have been conduct to
Unit (PCA-BMU) was proposed whereas the concept of build other version of Dataglove that share similar purpose
Euclidean Distance could be justify by the best grouping known as GloveMAP.
features according to the best neuron or winning neuron. The
conclusion will determine grasping features of subject to grasp There are a several types of object that involved and
the experimental object with the thumb, index and middle based on Cutkosky taxonomy, the lists of objects are as
fingers of GloveMAP. Based on fingers adapted grasping force follows; (a) Power Grip (Ball, Cylinder, Pen and Key), and
movement, this study gives the grasping features in order to (b) Precision Grip (Disc, Scissors, Pins and Paper).
justify the best grasping for each subject grasp behavior. According to Cutkosky [1], all subjects should confine to
single-handed operations and there should have been a
Index Term— Grasp feature, grasping force human better appreciation of how task requirements and object
grasping data, grasp behavior, and PCA-best matching unit geometry combine to justify the grasp choice for better
(PCA-BMU) result of human grasp. The next process flow is to eliminate
or minimize the unwanted signal and noise by using
I. INTRODUCTION Gaussian Filter. Gaussian Filtering makes grasping signal
THERE are too many applications in this era that related to become smoother and lessens the abrupt changes in signal
human gesture which is include parts of human body such as frequency. Then the grasping force signals are analyzed
hands, face, body and many more. Hand gesture is one of using PCA. Since PCA functions as data reduction, PCA
the famous gestures used in daily life. People use hand becomes the first choice method in reducing the redundancy
gesture to enhance the communication with others to deliver in grasping signal. PCA is capable to generate an
the information of thoughts effectively. This hand gesture “Eigenfinger” for thumb, index and middle fingers of
will give a lot of important information of fingers or hand grasping data.
movement that can be implemented in the industrial
applications such as video games industry, biomedical This research paper is structured as follows: Section 2
instrument, sports science, surveillance systems and many addresses the literature review of the related researches to
more. A device known as a Dataglove is presented as a the several approaches, applications and problems of
medium to measure the information gain from hand or recognizing the fingers grasping force signal. Section 3
fingers activities. Dataglove is known as cyberglove or describes the methodologies of the system. Section 4 will
wired glove and it is a device that can be donned by human. present the results and discussions. Finally on section 5
Any of physical data can be captured by this technology described the conclusions and proposing some possible
using various sensors such as bending sensor or mostly future work.
known as flexible bend sensor, force sensor or force
resistive sensor, tactile sensor and other types of sensor. II. LITERATURE REVIEW
There are numerous literatures on grasping force analysis
Nazrul H. ADNAN is with the Kolej Kemahiran Tinggi Mara (KKTM), and optimization developed over the last two decades.
Mukim Serom 4 & 5, 84410 Ledang, Johor Darul Takzim, MALAYSIA. Yoshikawa [2] introduced the concepts of active and passive
(corresponding author phone: +60194786589; fax: +6069756203; email: contact forces, and classified force closure into passive,
active and hybrid closures on the robotic hand especially for
[email protected]) the robotics hand grasp. The author also states to give the
Mahzan T. was with Kolej Kemahiran Tinggi Mara (KKTM), Mukim conditions for each types of force closure for a robotics
Serom 4 & 5, 84410 Ledang, Johor Darul Takzim, MALAYSIA, Phone: mechanism constrained. The effect of object weight on
grasping force has been investigated by Westling et al. [3].
+60192748040; email: [email protected]. The object surface friction plays an important role in
Khairunizam WAN was with 3Advanced Intelligent Computing and determining and controlling the grasping force, and it has
Sustainability Research Group, School of Mechatronic, Universiti been investigated by Howe et al. [4] and Tremblay et al. [5].
Malaysia Perlis Kampus Pauh Putra, 02600 Arau, Perlis, MALAYSIA,

Phone: +60192298030; email: [email protected]
Nabilah H.E was with 3Advanced Intelligent Computing and

Sustainability Research Group, School of Mechatronic, Universiti
Malaysia Perlis Kampus Pauh Putra, 02600 Arau, Perlis, MALAYSIA,

Phone: +60135903062; email: [email protected]

152206-8383-IJMME-IJENS © December 2015 IJENS IJENS

International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:15 No:06 60

Shunji SHIMIZU et al. [6] develop the sensory glove called its special kinematical structure. According to Radwin, R.
as MKIII for measuring the grasping force distribution G.et al. and Swanson B. et al. stated that thumb, index and
among the human grasp activities. The sensory glove was middle fingers give more grips and stronger compare to the
developed using sixteen sheets of sheet type pressure other fingers (ring and little fingers). Both researchers used
distribution sensor. many items in order to determine the best finger usage such
as chuck, pulp, and lateral pinches and these items are tested
For this research, GloveMAP was developed using three to 100 subjects [14][15]. Fig. 3 shows the example of
flexiforce sensors that attached at surface / palm of the fingers grasp testier and Fig. 4 shows a sample of human
thumb, index and middle fingers. Figure 1 shows the sample grasp object.
of flexiforce sensor that is intended for reading forces that
are perpendicular to the sensor plane. According to [7]
flexiforce sensor is suitable to be used for medical
compression bandages (MCB) to map the pressure applied
by compression products at multiple points. Flexiforce
sensor also capable to be function as haptic interface [8] and
visual sensing systems by a virtual feel [9][10][11].

In this research paper the Gaussian filtering method
application is for noise suppression application whereas that
the noise is smoothed out, at the same time the signal is also
distorted. Timothy Popkin et al. [12] used Gaussian filter for
solve the blurring of images and Gaussian filter capable to
produce high accuracy and at greatly reduced computational
cost compared to the traditional method. The example of
waveform to perform the Gaussian filter was shown in fig.
2. For the dimensional data reduction method for grasping
force feature classification is using principal component
analysis (PCA) method. PCA capable to quantize and
characterize the variance in hand / grasping posture of novel
transformation task [13].

Fig. 3. Experiment of fingers grasps [14]

Fig. 1. GloveMAP Dataglove and Flexiforce Sensor Fig. 4. Object grasping

Fig. 2. Gaussian Filter A. Gaussian Filtering Techniques

III. METHODOLOGY GloveMAP signal is prepared with Gaussian filtering
The measures of the three main finger movements are method in order to remove noise produced by random
well-defined in a marginally in different way of grasp due to thermal motion of charge inside the electrical conductor.
Noise within signal could affect the performance of objects’
feature and classification. Resistors used in GloveMAP also
would produce noise as heat inside resistors buildup. Each
data collection from 8 objects will be filtered using Gaussian
Filtering. Fig. 5(a) and Fig. 5(b) show unfiltered and filtered
voltage produced from human grasping. Both figures
demonstrate the result of Gaussian Filtering into raw voltage
to reduce noises and overshoot. Gaussian has an advantage
of reducing noises and overshoot of the input grasping
signal.

152206-8383-IJMME-IJENS © December 2015 IJENS IJENS

International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:15 No:06 61

converted to mass by dividing the values with 1000 to

change from gram to kilogram. Acceleration or gravity
needs to be considered due to the Newton’s law in finding

force value.

(2)

Where ,

F = force in Newton
m = mass in kilogram
a = acceleration ( gravity)

Fig. 5. a) Unfiltered voltage output with noise b) Voltage with Gaussian D. PCA-Best Matching Unit (BMU) Feature
Filter The Best Matching Unit feature is taken from the

B. Polynomial regression competitive learning of PCA as shown in Fig. 7. Based on
Fig. 4, the output of PCA, namely as the set of principal
The output signal of the force resistive sensor used in this components, are functioning as the input of BMU. The
research is in a voltage value which is in range of 0 to 5V. BMU objective is to cluster all data into a set of groups. The
These signals are need to be converted into a force value to clustering is also capable to separate the data which appear
give a better understanding about force applied to the similar, close to one another and place the very different
fingertips during grasping activities. The force resistive ones distant from one another. Suppose that the input y = [
sensor’s datasheet shows the weight (g) versus voltage (V)
graph as in Fig. 6. Based on the graph, the information of , ,….., ]T, the weight vector of the neuron j in BMU
weight (g) value can be determined. In order to find the is = [ , ,….., ]T.
relationship between weight and voltage, the polynomial
regression is used. The polynomial regression is one types Fig. 7. PCA and BMU
of regression analysis used to model a nonlinear relationship
between two different variables (independent x and
dependent y) to fit the nonlinear data and to describe its
phenomena. Fifth degree polynomial is determined as the
best data fit compared to the other polynomial equation such
as linear, quadratic and cubic.

The voltage output of the grasping object is used to
substitute into the x variable in the fifth degree polynomial
equation as in Equation 1 to determine the weight (g) value
(y variable).The information of weight (g) are then will be
used in the next step in order to convert the output voltage
signal into the output force signal.

(1)
IV. RESULT AND DISCUSSION

In this section, the analyses of overall step results are
started accordingly from data acquisition, data analysis,
features usage, and finally classifier recognition result.

Fig. 6. The 5th degree polynomial equation applied to the graph A. Experiment Setup

C. Force formula This experiment classify the grasping force of different
Force value can be calculated by using Newton’s second shapes of objects, about 10 human subjects are participated
in this experiment in order to obtain the information data
law, as in Equation (2). Based on the polynomial regression from the grasping activities. These objects were selected
applied to the graph of weight (g) versus voltage (V), weight based on the pattern grip that excessively used thumb, index
(g) data is determined. The values of weight (g) are then and middle fingers. The selected objects are placed on the
plane surface and the subjects are asked to grasp the objects
and hold it for about 3seconds. Fig. 8 shows how the signal
extracted using GloveMAP Dataglove.

B. Human Grasping Force Data

The signal from the force sensor that equipped to the
glove on to the thumb, index and middle fingers were form
in voltage output signal. The grasping objects activity was
done in 10 seconds to observe the output signal pattern. The
signal output show zero (0) voltage at 0-1.9s and 8.1 -10s

152206-8383-IJMME-IJENS © December 2015 IJENS IJENS

International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:15 No:06 62

due to the non-contact of fingers and object. Meanwhile, the on the explanation in the previous chapter, centroid or the
voltage signal at 2.0-8.0s show the information of the winning neuron is formed by the competition of each neuron
fingers produced from the grasping activity. Thumb fingers for representation of the group of data. The process of
indicated about 2.69-2.84 volt, the higher voltage from the competing occurs until a next competing between other
other fingers while the index and middle fingers indicated neuron except the centroid or winning neuron had been
about 2.5 -2.74 volt and 2.31-2.49 volt respectively. Fig. 9. finalized. To determine the PCA-Best Matching Unit (PCA-
shows the results of finger force for object cylinder. BMU) of the grasping data, the concept of neighborhood
between neuron was applied. So, the next step was
calculating all neurons nearby the centroid or winning
neuron as shown in Fig. 10. One method to calculate the
neighborhood between nodes and centroid or winning
neuron is the Euclidean distance. The steps to determine the
nearest node to the winning neuron or centroid are stated
below.

Fig. 8. GloveMAP data extracted for Ball and Dice Object 1) All nodes were justified using Euclidean Distance to
winning neuron or centroid.

2) The equivalent or nearest node matching with any of
the centroids were justified.

3) The nearest nodes to the winning neuron or centroid
will form a group of node identified as “Cluster”.

Fig. 9. The signal output pattern from the object grasping activities. Fig. 10 also shows the nodes that stay far away from the
BMU could be eliminated in order to get the best BMU
Table I neighborhood and at the same time BMU capable to help
The output transformation data in voltage into the force form. PCA form grasping force features. Figure 11 shows the
clustering grasping feature for PCA-BMU approach. The
figures show that the outer data (stated in the figure 10)
could be eliminated or the data range could be reduced. The
reduction / eliminate process should be considered because
of the grasping force feature itself, could be more accurate
and always emerges. The figures also show the 2 groups of
grasping force. Group 1 shows the maximum data features
compare to the Group 2. Meaning that whichever the
clustering group shows the maximum dataset, the selected
clustering group could be called as the grasping force
feature.

Fig. 10. The BMU of finger grasping (a) Ball (b) Glass (c) Cylinder (d)
Dice

C. PCA-Best Matching Unit (PCA-BMU)

The process of justifying the human grasping data
involves several grasping groups and forming a component
identified as nodes or neuron. The group of neurons
basically has one main neuron located at the center of group
of neurons, which is the winning neuron or centroid. Based

152206-8383-IJMME-IJENS © December 2015 IJENS IJENS

International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:15 No:06 63

Fig. 11. PCA-BMU data clustering for fingers grasping (a) Ball (b) Glass Force Distribution Pattern in Grasping. in ICAR IEEE, pp. 299-304,
(c) Cylinder (d) Dice 1994,DOI: 10.1109/ICAR.1997.620198
[7] Jawad Al Khaburi, Abbas A. Dehghani-Sanij, E. Andrea Nelson and
V. CONCLUSION AND FUTURE WORKS Jerry Hutchinson. Pressure Mapping Bandage Prototype:
In this paper, we proposed the method to classify Development and Testing. In International Conference on Biomedical
fingertip grasping force signal for several selected objects Engineering (ICoBE), 430-435, 2012.
which is based on Polynomial Regression and PCA-BMU [8] Zhengmao Ye and Gregory Auner. Haptic Interface Prototype for
techniques. The chosen of Best Matching Unit (BMU) for Feedback Control on Robotic Integration of Smart Sensors. In IEEE,
this research capable to generate the best method to 995-1000, 2003, DOI: 10.1109/CCA.2003.1223146
smoothen up the grasp signal of grasping features without [9] J. E. Colgate and G. G. Schenkel. Passivity of a Class of Sampled-
any noise disturbance. Experimental results show that the Data System: Application to Haptic interface. Journal of Robotic
both method works well in defining grasping force types Systems 14(1), 31-41, 1997.
with only a usage of few principal components and also [10] A. Hannaford and J. Ryu. Time-Domain Passivity Control of Haptic
capable to identifying the grasp type of an input motion Interfaces. In IEEE Transactions on Robotics and Automation, Vol.
data. For future plan works, the results are by adding the 18, No. 1, I-10, 2002, DOI: 10.1109/70.988969
signal processing technique in the research. This signal [11] B. E. Miller, J. E. Colgate and R.A Freeman. Guaranteed Stability of
processing will be employed in solving the problem of Haptic System with Nonlinear Virtual Environments. In IEEE
analysing more sophisticated signal pattern especially on the Transactions on Robotics and Automation, Vol. 16, No. 6, 712-719,
signal produced during transition gesture and continuing 2000, DOI: 10.1109/70.897782
gesture [12] Timothy Popkin, Andrea Cavallaro, and David Hands. Accurate and
Efficient Method for Smoothly Space-Variant Gaussian Blurring. In
ACKNOWLEDGMENT IEEE Transactions On Image Processing, Vol. 19, No. 5, pp. 1362-
Special thanks to all members of KKTM Ledang and 1370, 2010, DOI: 10.1109/TIP.2010.2041400
MARA for providing the research equipment’s and internal [13] Ramana Vinjamuri, Mingui Sun, Douglas Weber, Wei Wang, Donald
foundations. This work was supported by the Majlis Crammond and Zhi-Hong Mao. Quantizing and Characterizing the
Amanah Rakyat (MARA). Variance of Hand Postures in a Novel Transformation Task. In 31st
Annual International Conference, 5312-5315, 2009.
REFERENCES [14] Radwin, R. G., Oh, S., Jensen, T. R., & Webster, J. G. (1992).
External finger forces in submaximal five-finger static pinch
[1] Cutkosky, M. R. (1989). On grasp choice, grasp models, and the prehension. Ergonomics, 35(3), 275-288.
design of hands for manufacturing tasks. Robotics and Automation, [15] Swanson, A. B., Matev, I. B., & De Groot, G. (1970). The strength of
IEEE Transactions on, 5(3), 269-279. the hand.Bull Prosthet Res, 10(14), 145-153.

[2] T.Yoshikawa. Passive and active closures by constraining Nazrul H. ADNAN received his Bachelor Engineering (Hons) in Power
mechanisms. In Proceedings of IEEE International Conference on Electrical from Universiti Teknologi MARA (UiTM) and Master
Robotics and Automation, pages 1477–1482, 1996, DOI: Engineering in Advanced Manufacturing Technology from Universiti
10.1109/ROBOT.1996.506914 Teknologi Malaysia (UTM) since 2004 and 2010 respectively. He was
awarded Ph.D from Universiti Malaysia Perlis (UniMAP) in 2015 where
[3] Westling, G. and R. S. Johansson. Factors Influencing the Force his Ph.D thesis interest is in Biomedical, BioScience, Human-Computer
Control During Precision Grip. Experimental Brain Research, volume Interaction (HCI), Product Design, Artificial Intelligence, Signal Processing
513, pp. 277- 284, 1984. and Machine Design.
Mahzan T. received his Bachelor Electrical Engineering (Hons) in Control
[4] Howe, Robert D., Nicolas Popp, P. Akella, Imin Kao and M. & Instrumentation from University Technology Malaysia (UTM) and
Cutkosky. Grasping, manipulation and control with tactile sensing. Master Technical Education in Industrial Electronics from University Tun
Proceedings of the IEEE International Conference on Robotics and Hussein Onn Malaysia (UTHM) since 1997 and 2007 respectively.
Automation. pp. 1258-1263,1990, DOI: Currently he is Director of Kolej Kemahiran Tinggi Mara Ledang, Johor
10.1109/ROBOT.1990.126171 whose responsible for the strategic positioning and pursuit of strategic
goals as derived by MARA.
[5] Tremblay, Marc R. and Mark R. Cutkosky. Estimating friction using Khairunizam WAN received his B. Eng. degree in Electrical & Electronic
incipient slip sensing during a manipulation task. Proceedings of the Eng. from Yamaguchi University and Ph.D. in Mechatronic Eng. from
IEEE International Conference on Robotics and Automation, pp. 429- Kagawa University, in 1999 and 2009 respectively. He is currently a Senior
434,1993, DOI: 10.1109/ROBOT.1993.292018 Lecturer at School Of Mechatronic Engineering, University Malaysia
Perlis. He is member of Board of Engineer and Institute of Engineer,
[6] Shunji SHIMIZU, Makoto SHIMOJO, Sigeru SATO, Yoshikazu Malaysia. His research interest is in Human-Computer Interaction (HCI),
SEKI, Akihiko TAKAHASHI, Yukio INUKAI and Matsutaro Intelligent Transportation System, Artificial Intelligence and Robotics.
YOSHIOKA. The Relationship between Human Grip Types and Nabilah H.E received her B.Eng. degree in Mechatronic Engineering from
UniMAP and now was in final year for Master Degree also in Mechatronic
Engineering from UniMAP.

152206-8383-IJMME-IJENS © December 2015 IJENS IJENS

2010 IEEE EMBS Conference on Biomedical Engineering & Sciences (IECBES 2010), Kuala Lumpur, Malaysia, 30th November - 2nd December 2010.

Initial Results on Magnetic Induction
Tomography Hardware Measurement Using Hall

Effect Sensor Application

Zulkarnay Zakaria1,2, Noor Aqma Hj. Mohd Yazid2, Noor Hidayah Jaafar2, Mohd Fahajumi Jumaah2, Muhammad
Saiful Badri Mansor2 and Ruzairi Abdul Rahim3

1 Tomography Imaging Research Group, School of Mechatronic Engineering
Universiti Malaysia Perlis, 02600 Jejawi, Arau, Perlis, MALAYSIA
[email protected]

2 Biomedical Electronic Engineering, School of Mechatronic Engineering
Universiti Malaysia Perlis, 02600 Jejawi, Arau, Perlis, MALAYSIA
[email protected]
[email protected]
[email protected]
[email protected]

3Process Tomography Research Group (PROTOM), Control and Instrumentation Engineering Department,
Faculty of Electrical Engineering, Universiti Teknologi Malaysia,
81310 Skudai, Johor, Malaysia.
[email protected]

Abstract - Magnetic induction tomography (MIT) is instrumentation [1]. Originated from the Greek words
among the new technology that compliment other ‘tomos’ which means slice and ‘graph’ meaning picture,
tomography methods such as ultrasonic, optical, tomography can be defined as a picture of a slice . In simple
capacitance and several others. This type of tomography terms, tomography is an imaging technique that enables one
applies the magnetic field to detect the existence of the to determine the contents of a closed system without
object that is going to image. Several methods are physically looking inside it [2].
possible in constructing the magnetic induction
hardware. Most of the researchers used coils for both There are different requirements in an industrial
transmitter and receiver which are more complicated environment than there are within a medical one: different
and need large space. The Hall Effect sensors (HES) regulations regarding for example use of ionizing modalities
have the potential of replacing the coil at the receiver and different speed requirements [3]. Technically, Process
side since it has the ability to measure the value of Tomography can be described as imaging process
magnetic field strength and convert it to voltage value. parameters in space and time. Important flow information
This concept is same as the ultrasonic sensors used in such as concentration measurement, velocity, flow rate, flow
compositions and others can be obtained without the need to
ultrasonic tomography instrumentation hardware. The invade the process or object.
results have shown that the pattern of capture
data by hall effect sensor are almost the same As a result, cross sectional images of processes
pattern for all 8 sensors for each material used. generate better online inspection, monitoring and process
control - promoting improved yields and more effective
This have given positive sign that HES is capable to be utilization of available process capacity. Potentially,
applied in MIT measurement system. tomographic systems may also be an alternative approach in
. developing and verifying process theories and models, as
well as for improving process instrumentation [4].
Keywords- Magnetic Induction, Hall Effect sensor,
magnetic field, tomography, inductor In MIT, the signal detected by the sensor coil is
composed of two components, the primary and secondary
I. INTRODUCTION signals. The primary signal is due to direct induction by the
field from the excitation coil. The secondary signal is due to
The widespread need for direct analysis of the the eddy currents induced in the sample which in turn create
internal characteristics of process plants in order to improve their own magnetic field. The secondary signal is very small
the design and operation of equipment has made process in magnitude, typically only 1% of the primary signal for an
tomography a main research activity within the industrial operating frequency of 10 MHz [4], [5], [6].

Magnetic field characteristic is used to measure the
magnetic induction that generates by the sample itself and
also the coils .Therefore it needs to produce the magnetic

978-1-4244-7600-8/10/$26.00 ©2010 IEEE 9

field. There are mainly two type of magnetic source that But the disadvantage is that a given gradiometer
could be used. First is a permanent magnet. The permanent can only cancel out the primary signal for one excitation coil
magnet can be various size, shape and strength. It depends position. Another approach is to use an array of single coils
on the direction f the magnetic field would be measure. as sensors with a fully electronic back off [8], [9].
Second is current inducing magnetic field using coil. The
coil produce magnetic field when the current flow. In this II. METHODOLOGY
project, coil use instead of permanent magnet. The reason is This project required hardware as signal
that magnetic field produce by coil can easily controlled in conditioning circuit that is connected to microcontroller. The
term of direction, strength and time. microcontroller will control the signal projection timing for
generated at four coils while eight HESs are placed at
In this project, eight magnetic sensor is use to detect equidistance around the sensor jig as in Fig. 1. Conditioning
the magnetic field. The magnetic sensor use is Hall Effect circuit is used to remove the noise from the sensor and to
type magnetic sensor with ratio metric output type. The amplify the signal.
sensor is placed around the sensor jig. It will detect magnetic The data is next will be captured using Labview
field created by coils. The magnetic field is detected is in software and then will be analyzed.
vertical direction. The coils and the sensor are controlled by A. Hardware development
an embedded system. Basically the embedded system The hardware consists several parts that are signal
activates the coil and takes measurement of the sensor. conditioning circuit and sensor jig which contains HESs
and inductor as in Fig. 1, Fig. 2 and Fig. 3.
Currently magnetic induction tomography (MIT)
applies magnetic field to detect the object’s electrical Fig. 1 Sensor jig development
conductivity which is generated by a coil that act as a
transmitter. The objects can be a biological tissue or fluid or Fig. 2 Magnetic Induction Tomography Hardware System
metals. The object that feel this magnetic field then will
produce its own field which is known as secondary field or
eddy current which will be detected by the receiver which is
also a coil. The application of coils as the receiver will need a
larger space since the size of the coil itself is relatively much
bigger compare to hall effect sensor (HES). Other than that,
through the application of coils at the receiver side, the
image reconstruction algorithm is quite complex since it is
involve finite element technique in identifying the current
flows in the coils at the receiver. In tomography techniques,
normally the application of more receivers will produce a
better quality of the reconstructed images. Due to this and
the complexity of image reconstruction algorithm by
applying the coils at the receiver, this research is carried out
in studying and understanding the potential role of hall effect
sensors application in magnetic induction tomography
system in replacing the coils at the receivers.

Magnetic induction tomography is used to image
the electrical properties inside a region of interest. The
systems differ in the construction of the receiver channels
which can be composed of coils or gradiometers. Excitation
coils are used to generate a primary magnetic field and the
secondary magnetic field, which is caused by the eddy
currents, includes the information about the conductivity
distribution inside the object [6], [7].

A difficulty is that the secondary field due to the
eddy currents inside the object is superimposed on the
primary magnetic field. But the primary magnetic field is
much larger (factor 102–106) than the secondary magnetic
field depending on the frequency and coil geometry. To
improve the sensitivity of the measurement, a separate
sensing coil can be used for the subtraction of the primary
signal [1], [3].

Another method is the overlapping of excitation
and sensing coils described by Peyton et al (1999), but this is
effective only for the sensing coils immediately adjacent to
the excitation coil. Use a planar gradiometer as the sensor
which can be made mechanically very stable and most of the
primary signal can be cancelled out [2], [4].

10

Fig. 3 The sensor jig with the phantom In this project, the power source for the sensor is 9v where it
can withstand voltage up to 11v. The zero magnetic field
B. Electromagnetic Coil input would result output of 4.5v from the sensor.

Magnetic inductor is made of wrapping wire. The D. Signal conditioning circuit
wire had a conducting size of AWG 30.The conditioning
material is made of cooper. The diameter of the wire is The signal conditioning circuit uses two operational
0.254mm while the resistance 338.496 ohm per kilometer amplifiers and one instrument amplifier. The operational
with maximum current 3A. amplifier is use as voltage follower to provide high
impedance of signal and removing any negative voltage
from the signal. The operational amplifier is use to offset the
signal from the noise and to amplify the signal.

E. Data Acquisition Systems (DAQ)

For this project, we used NI USB-6009 14-Bit, 48 kS/s
Low-Cost Multifunction DAQ model from National
Instruments. It offers direct control of all hardware on the
DAQ board from the LabVIEW software.

Fig. 4 Inductor using in the project Fig. 6 NI USB-6009 hardware

C. Hall Effect Sensor (HES) III. RESULTS AND DISCUSSIONS

The sensor use on the project is Hall effect sensors In this experiment, three samples have been taken as the
as in Fig. 2. It is used to detect the strength of the magnetic object which are NaCl solution, Engine oil an also biological
field that had been produce by the coil [10]. Sensitivity of the tissue (beef) as shown in Fig. 7. The results have shown that
sensor is 3.125mV/Gauss. Output of the sensor is ratio the pattern of data are almost the same but with different
metric type. Gaussian range for the sensor to be able to magnitude.
detect is between 0-640 Gauss in either direction. When
there is no magnetic field detected, the sensor give a constant
output 2.5v using a power source of 5V.

Fig. 5 Hall Effect Sensor SS94A2 Fig. 7 Hall effect sensor (HES) measurement values in mV for three
different type of materials

11

IV. CONCLUSION & DISCUSSION

A new approach of magnetic induction tomography sensor
jig design has been produced. This new hardware system has
been tested with three different samples including biological
soft tissues and the initial data have shown that the
developed system is capable of measuring the magnetic field
due to NaCl, Engine oil and biological tissue. As conclusion
we can say that this device has shown its functionality and
have the potential in both industrial processes an also in
medical applications through the use of hall effect sensors.

ACKNOWLEDGMENT

This work is supported by the FRGS grant 9003-00248 by
Ministry of Higher Education, Malaysia.

REFERENCES

[1] Robert Merwa and Hermann Scharfetter “Magnetic induction

tomography: comparison of the image quality using different types of
Receivers”, Physiological Measurement, Vol. 29, Issue 6, pp. 417- 429,

7 Dec 2007.

[2] Xu Li, Yuan Xu, and Bin Hea ,”Magnetoacoustic tomography with

magnetic induction for imaging electrical impedance of biological

tissue“,Journal of Applied Physics, Vol. 99, Issue 6, pp. 066112-

066112-3, Mar 2006.

[3] Zheng Xu, Haijun Luo, Wei He, Chuanhong He, Xiaodong Song and

Zhanglong Zahng “A multi-channel magnetic induction tomography

measurement system for human brain model imaging” , Physiological

Measurement, Vol. 30, Issue 6, pp. 175- 186, 29 Nov 2008.

[4] A. Martinez Olmos , J. Alberdi Primicia and J. L. Fernandez

Marron ,”Influence of Shielding Arrangement on ECT” Sensors, Vol.

6, Issue 9, pp. 1118-1127, Sept 2006.

[5] Ruzairi Abdul Rahim , Mohd Hafiz Fazalul Rahiman , Leong Lai

Chen , Chan Kok San and Pang Jon Fea “Hardware Implementation of

Multiple Fan Beam Projection Technique in Optical Fibre Process

Tomography” Sensors, Vol. 8, Issue 5, pp. 3406-3428, March 2008.
[6] Chen, D., Pan, M., Luo, F., “Study on Accurate 3D Magnetic Field

Measurement System”, The Eighth International Conference on

Electronic Measurement and Instruments ICEMI’2007, Vol. 2, pp.

680-683.

[7] C. C. H. Lo, J. A. Paulsen, and D. C. Jiles, “A Magnetic Imaging

System for Evaluation of Material Conditions Using Magnetoresistive

Devices”, IEEE Transactions on Magnetics, Vol. 39, No. 5, pp.

3453-3455, September 2003

[8] S. Watson, R. J. Williams, H. Griffiths, W. Gough and A. Morris, “A

Transceiver For Direct Phase Measurement Magnetic Induction

Tomography”, 23rd Annual International Conference of the IEEE

Engineering in Medicine and Biology Society, Vol. 4, pp. 3182-3184,

October 25-28, 2001, Istanbul, Turkey

[9] Ostovic, V., “A Simplified Approach to Magnetic Equivalent-Circuit

Modeling of induction Machines”, IEEE Transactions On Industry

Applications, Vol. 24. No. 2, pp. 308-316, April 1988.

[10] Carter, N.P., Ferrera, S., Kothari, L., Ye, S., “Hall-Effect Circuits and

Architectures for Non-Volatile System Design”, Proceedings of the

2005 European Conference on Circuit Theory and Design, 2005, Vol.

2, pp. 131-134.

12

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

DEVELOPMENT OF RED BLOOD CELL ANALYSIS SYSTEM USING
NI VISION BUILDER AI

Razali Tomari1, Jalil Lias2, Rabiatuladawiah Musa2 and Wan Nurshazwani Wan Zakaria1

1Advanced Mechatronic Research Group (ADMIRE), Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn
Malaysia, Parit Raja Batu Pahat, Malaysia

2Department of Biomedical Electronics Engineering, Kolej Kemahiran Tinggi MARA Ledang, Serom 4 and 5, Jalan Serom-Bukit
Gambir, Sungai Mati, Ledang, Johor, Malaysia

E-Mail: [email protected]

ABSTRACT

Red blood cell (RBC) diagnosis is very important process for early detection of blood related disease such as
malaria and anemia before suitable follow up treatment can be proceed. Conventional method is conducted by pathologist
by manually count and classifies the viewed cell under light microscope. Such process is tedious and required highly skill
and experience pathologist to analyze the shape of the red blood cell and consequently counting its number. In this paper
an automated RBC counting and classification system is proposed by using National Instrument (NI) Vision Builder
Automated Inspection (AI) tool to speed up the time consumption to analyze the RBC and to reduce the potential of the
wrongly identified RBC. Initially the RBC image undergoes image pre-processing steps which involved global threshold of
method applied green channel color image. Then it continues with RBC counting by using particle area and calculator
numeric function method. Eventually, Heywood Circularity Factor method is applied for normal and abnormal RBC
classification. The proposed method has been tested on blood cell images and the effectiveness and reliability of the system
has been demonstrated.

Keywords: red blood cell, NI vision builder AI, particle area, heywood circularity factor.

INTRODUCTION Image processing is an alternative method to
identify each single cell in blood samples. Comparing to
Blood is a connective tissue consisting of red the conventional method, image processing provide
blood cells (RBC), white blood cells (WBC), and platelets various advantages that y helpful in the RBC analysis.
suspended in plasma. As a medium of transportation for Classification of each cell in blood samples is the ultimate
the whole body, blood composition is very vital to be process to be conducted to identify each single cell in the
monitored when it comes to medical inspection. RBC blood cell. This classification process will lead to the
analysis contributes information of pathological diseases counting process. From counting process the total quantity
and condition. It helps doctors to determine the of each single cell in blood samples can be gained. It will
appropriate treatment to the patient. Any condition which be used as a reference for a doctor to make a conclusion
there is an abnormally low of haemoglobin concentration about the patient health status.
or red blood cell count is indicating to anaemia (Fox,
2009) and also low of specific vitamin (Sharif, 2012). The Figure-1. Image of blood sample.
shape of RBC and its deformability or abnormality has
connection to the relevant disease such as Huntington’s
disease, Myalgic Encephalomyelitis (ME) and Multiple
Sclerosis (MS) (Vromen, 2009).

Complete blood count (CBC) is a compilation
test of blood component including for RBC analysis.
Normally the blood sample is taken and processed in the
laboratory by using chemical electronic devices such as
haemocytometer or haematology analyzer. This work is
manually done by lab technologist and it is very dependent
to their skill especially to count the cell through
microscope (Sharif, 2012). The counting and analysis
process is difficult when the cells are overlapped as shown
in Figure-1. The conventional method is time-consuming
to complete and the counting task is laborious (Natsution,
2008).

8692

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Nowadays many researchers have interest in to produce strong analysis approach for medical diagnosis
developing algorithm to automatically count and analyze purpose.
the blood cells. They applied advanced technology of SYSTEM OVERVIEW
image processing in conjunction with artificial intelligent
and computer graphic system to provide methods that are The proposed method flow diagram of RBC
accurate, faster and easy to use. Source of blood samples counting and classification system is outlined in Figure-2.
are obtained from the lab and the images are captured It starts with the input image acquired from a light
using light microscope that attached with a digital camera microscope that attached with eye piece static camera. The
(Tomari, 2014; Venkatalaksmi, 2013). For the image pre-processing start with global threshold RGB colour
processing part, many advance software are used such as under green color channel. The range of the value is finely
MATLAB (sharif, 2012), Lab VIEW (Ajay, 2014) selected to extract only the blood cells from its
Microsoft Visual Studio and Open CV (Tomari, 2015). background. After that image smoothing process continues
The image processing can be broadly divided into two with remove border object, remove small object and fill
main categories which are segmentation and classification. the hole. Once the pre-processing step completed, the
The former extract the RBC area while the latter process counting process starts by using selection range of particle
the extracted area into cell morphology information. area to classify single, two-overlap, three- overlap of RBC
and so on. Following that, the calculator numeric function
Segmentation process is an important step in is used to sum up all of the RBC quantity.
image processing to differentiate objects from the
background. In the previous study, there are variety of For classification, since the unique feature
methods proposed for segmenting the RBC image such as between the normal and abnormal RBC is the spike
threshold and morphological operation (Ajay, 2014, border, hence Heywood circularity factor can be used to
Venkatalakshmi, 2013; Berge, 2011), pulse coupled neural differentiate between them. The significant value of
network (PCNN) (Adagale, 2013), colour based Heywood circularity is finely tuned until it can
segmentation (Mahmood, 2013), masking and watershed significantly distinguish between the two cells. For this
algorithm (sharif, 2012), active appearance model (Cai, project, all of the mentioned methods are applied using NI
2012), shape reconstruction (Wang, 2010) and region Vision Builder AI. Towards the end, the performances and
based segmentation (Vromen, 2009). In this paper, RBCs accuracy of this counting and classification method are
image is segmented using threshold and morphological evaluated.
operation. The value of threshold is set in green channel
since it gives the best contrast between RBC and Figure-2. RBC counting and classification flow diagram.
background (Tomari, 2014). Morphological operation then
works on binary image for changing shape, size, structure,
and extract clear feature of RBCs. It includes remove
border, remove small object, fill holes, reverse and
equalize steps.

Classification of RBC process can be carried out
using several methods such as neural network (Tomari,
2014, Poomcokrak, 2008), template matching (Adagale,
2013), edge detection (Cai, 2012), and depth map and
surface feature (Wang, 2008). This process is to identify or
recognize the pattern of RBC into normal and abnormal
considering single or overlapped condition. Besides, it also
aids in counting process. This paper uses Heywood
circularity factor technique for RBC classification added
with mathematical numeric function for calculation
purpose.

RBC counting and analysis using image
processing is ongoing research in medical diagnosis
imaging. It always goes through improvement as a new
research is carried out. The researchers offered many
different methods that can provide better accuracy and
promising solution. However there are still weaknesses
and constraints due to the image itself such as colour
similarity, weak edge boundary, overlapping condition,
image quality, contrast, brightness, illumination and noise.
Thus, more studies must be done to handle those matters

8693

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Image acquisition

The images are acquired from light microscope
that equipped with DinoEye Eyepiece Camera as shown in
Figure-3, and the process of capturing the image will
involve blood smear process to the prepared sample.
Blood smear is a process of preparation blood specimen on
the slide that observed under microscope. The process for
displaying the RBC image will involve digitization of
image from the optical image with 40 times (40X)
objective which equal to approximately 400 magnification.

Figure-3. Image acquisition equipment.

Image segmentation and processing Figure-4. Image pre-processing stages.

Before proceed for RBC identification, the initial The threshold value can be selected between 0
stage is pre-process the acquired image. This image cannot until 255 and among this value the most significant
be process straight away as it is. It needs to be prepared in threshold that matches for the whole sample images is in
monochrome or binary representation for convenience and the range of 148-154. There are total four images that be
fast processing. Binary image represents the image as 1 or used in this project for RBC counting and classification
0 and from this representation then the next process can be and sample of the threshold image is shown in the first
proceed. Threshold method is the easiest way to separate row of Figure-4. In the image, the background is in the
the object in image from its background. The color white representation while the foreground is in black.
property that is selected for this project is RGB. Under
RGB segregation, the most prominent color to After the thresholding process, the next process is
qualitatively and quantitatively differentiate between RBC applying the Vision Assistant function to pre-process the
and its background is the Green channel. image. Due to the process only focusing on the complete
RBC shape, thus the RBC in the border location can be
ignored due to its incomplete shape. Under the tab of
“Processing Function: Binary” by selecting “Advanced
Morphology” function follow by “Remove border object”
option, the RBC that located in the image border can be
removed. Sample of the outcome of this process is shown
in the second row of Figure-4.

From the segmentation process there are small
objects and holes exist in the image that needs to be
cleared. These small objects are come from the platelet

8694

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

area and from the imperfection of background EXPERIMENTAL RESULT AND DISCUSSION
identification. To diminish such noise a “Remove small
object” function is selected and minimum area that In this section, analysis of the developed RBC
constitute to noise is empirically determined as 50. Due to counting and classification system is tested using four
blood smeared images as shown in Figure-4 which are
hemoglobin shape of the RBC, after the threshold process labeled as RBC_1, RBC_2, RBC_3 and RBC_4.The
some of the inner RBC area becomes hollow. These holes analysis is conducted to determine an optimal parameters
and consequently the system performance for
need to be filled due to it affects in the classification stage distinguishing between single and overlap RBC, and
normal and abnormal RBC.
that using object particle area property. To solve this
matter, a “Fill hole” function is selected under the
“Processing Function: Binary” tab of Vision Assistant
function. Sample of the output of the filtering step is
shown in the third and fourth row of Figure-4.The image is

then ready to be fed into the next stage which are RBC

classification and counting.

RBC counting and classification (RBC_1) (RBC_2)

To determine the RBC morphology, in this (RBC_3) (RBC_4)
project two features are extracted which are the area of
particle and Heywood circularity. The first feature can
determine single and overlap RBC while the latter classify
the single RBC region between normal and abnormal. In
NI Vision Builder this method is located under “Detect
Objects” function under the tab of “Inspection Steps:
Locate Feature”. From empirically analysis, it is found
that the single non overlap RBC particle area is around
(524-3047) thus for area higher than this value the RBC
can be classified as overlap. The single RBC can be count
directly and percentage error count can be determined by
using following equation.

% − . | × % (1) Figure-5. Four samples of RBC image.
= | . .
From the analysis of four sample images shown
Heywood circularity factor function is applied to in Figure-5, it is found that the range of particle area of
identify the normal RBC region. It works by dividing the RBC in the images is from 574-16841 (RBC_1), 528-
object perimeter with the circumference of a circle with 15905 (RBC_2), 814-30838 (RBC_3), and 1352-7196
the same area. The closer the shape of a particle is to a (RBC_4). From this range, empirically the single RBC
disk, the closer the Heywood circularity factor is to 1. The region can be identified when the area is within 524 to
equation for calculating the factor is as follow: 3047 pixels. Above this value the RBC is identified as an
overlap cell.

In the overlap case, generally there will be a
= (2) situation where more than two RBCs in overlapped region
ℎ and hence the exact number is unknown. To solve this
issue, the range particle area is further manipulated. The
= √ range is set according to following formulation : 0-3047
for single RBC, 3048- 5000 for two RBCs, 5001-7000 for
three RBCs, 7001-9000 for four RBCs, 9001-11000 for
five RBCs and 11001-20000 is for six RBCs. With this
Once the normal RBC area is found, the simple procedure, the estimated number of total RBCs in
abnormal RBC location can be identified by subtracting the blood smeared image can be determine easily with an
acceptable accuracy.
the area of non overlap RBC with the normal RBC regions
For analysis of normal and abnormal RBCs from
as in equation (3). This method function is located in the single RBC cells, Heywood circularity factor function
“Calculator” function under the tab of “Inspection Steps: is applied. From the mentioned four sample images above,
Use Additional Tools”. preliminary testing indicate that the range of Heywood
circularity factor for normal RBC is 1.029-1.964 (RBC_1),
Abnormal RBC = Non_overlap RBC - Normal RBC (3)

8695

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

1.008-2.221(RBC_2), 1.004-2.757 (RBC_3) and 1.002- Table-1 summarized the result of RBC
1.723 (RBC_4). For generalizing this finding, the value is classification. Overall the proposed method performs quite
set within 1-1.026 for normal RBC region and abnormal well with an average accuracy of 86.076%, average
beyond this range. The number of each respective cluster normal precision of 64.820% and 63.881% of average
will be counted after the identification process completed. recall. This means that most of the object classes are
correctly identified with an acceptable error rate. Actually
the classification process only focused on normal RBC
identification and the abnormal one is identified based on
the subtraction function. Therefore, the quantity of
abnormal RBC is highly depending on the correctness of
the normal RBC detection. It can be seen that in image
RBC_ 4, there are quite a lot FP of abnormal RBC
detected. This misclassification occurs due to the shape of
the RBC that is in the form of oval shape. Therefore it is
out of the range of the defined normal RBC shape. If the
maximum range of Heywood circularity factor is changed
from 1.026 to 1.030 some of this normal RBC can be
classified, but unfortunately the number of misclassified
abnormal as normal become higher.

Table-1. Result for identifying and count the number of
normal and abnormal RBC.

Figure-6. Image RBC_4 classification result. Figure-7 shows sample of counting result of image
Figure-7. Image RBC_4 counting result.
Figure-6 shows sample of the classification and
counting result of image RBC_4 by using the developed
system. From the image it can be seen that the total
quantity of the three clusters which are single RBC,
overlap RBC, normal RBC and abnormal RBC is
successfully obtained. The general system performance is
assed based on the system ability to correctly identify and
count the number of normal RBC cell and abnormal RBC
cell. For each images, the quantitative measurement is
performed based on True Positive (TP), False Positive
(FP), True Negative (TN), and False Negative (FN)
parameters. All the related formulas are shown in equation
(4). Precision provides information about how many of the
detected fraction cells are correct, and recall tells how
many cells are correctly detected in each class from the
whole image. The accuracy on the other hand, gives
evaluation about how well the overall system performance
with respect to the ground truth data

Precision = TP ; Recall = TP ; Accuracy =
TP+FP TP+FN
TP+TN
(4)
TP+FP+TN+FN

8696

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

RBC_4. In the image processing, the objects that is limitation of using Heywood circularity factor method
touch border are neglected. For obtaining the better result for oval shape of normal RBC.
of counting as well as the accuracy, the threshold value in
earlier process is very important. Overall counting In future it is suggested that the classification
performance is summarized in Table-2. It shows the focused more than Heywood circularity factor which can
overall error rate for every sample is less that 10% and identify the unique feature that only belong to the normal
hence it is quite acceptable. Image RBC_4 has less error and abnormal RBC. Thus, both classification of normal
since the number of single RBC is dominant and only RBC and abnormal RBC is independent and can generate
three group of overlapped RBC is in it. For image RBC_1, the accurate result of the total quantity of non-overlap
RBC_2 and RBC_3, the overlapped cells number are quite normal and abnormal RBC. It is also suggested to use
high and with the existing of white blood cell and platelet Particle Classification method under Vision Assistants
make the counting and the classification task quite function due to its function can train samples of images to
difficult. The error percentage of image RBC_1 is highest be classified.
because of the condition of overlapped cells is very
complicated. It is quite hard to exactly put the number In the other hand, during image pre-processing
because of the size and the shapes of the overlapped cells stage, the image of white blood cell is proposed to be
are different for each group. removed earlier for better RBC counting and classification
result. Method that is suggested for this purpose is by
Table-2. Error percentage of RBC counting. using CMYK color threshold segregation function. In
future if more image of RBC samples can be provided
CONCLUSIONS then the performance of the system can be tested more
In this paper a method of RBC counting and thoroughly for multi condition and shape of RBC.

classification is proposed. The system consists of two ACKNOWLEDGMENT
main process which is image pre-processing and RBC The authors would like to thank to Ministry of
counting and classification. The platform that is used is NI
Vision Builder AI. The main function which is applied in Education (MOE) and Universiti Tun Hussein onn
this project in the software is “Vision Assistant” under the Malaysia (UTHM) for supporting this research under
tab of “Inspection Step: Enhance Images and “Detect Research Acculturation Collaborative Effort (Vot. no.
Object” under the tab of “Inspection Steps: Locate 1448).
Features”. Vision Assistant if for image pre-processing
and Detect Object is for counting and classification. REFERENCES

We have shown that by using area of particle Ajay, P. Dhawale and Hirekhan, S. R. 2014. Real- Time
function, each of single RBC can be identified. Particle Image Processing for Biological Applications Through
area specialized applied for non-overlap RBC quantity. Morphological Operations using LabVIEW. International
For overlap RBC, there is additional method that was Journal of Engineering research and Technology (IJERT),
tested which is selection the range of particle for two 3(5).
overlap RBC, three overlap RBC, four overlap RBC and
so on. It is noted that selection particle area range is not Adagale, S.S. and Pawar, S.S. 2013. Image Segmentation
accurate and not robust for instance in RBC total overlap using PCNN and Template Matching for Blood Cell
case. In future for counting process, it suggested to apply Counting. IEEE International Conference on
Hough transform due to it can detect circle image in Computational Intelligence and Computing Research, pp.
overlap region. NI Vision Builder AI is not supported for 1-5.
this function.
Berge, H., Taylor, D., Krishnan, S. and Douglas, T.S.
Classification of normal and abnormal RBC was 2011. Improved Red Blood Cells Counting in Thin Blood
defined by using Heywood circularity factor. The tested Smear. IEEE- ISBI, pp. 204-207.
system was successfully done with the acceptable
precision, recall and accuracy of the normal and abnormal Cai, R., Wu, Q., Zhang, R., Fan, L. and Ruan, C. 2012.
RBC if compared to the ground truth data. In this project Red Blood Cell Segmentation using Active Appearance
the classification only focused on normal RBC and the Model. IEEE 11th International Conference on Signal
abnormal RBC data is not independent. By the way, there Processing, 3, pp. 1641-1644.

Fox, S.I. 2009. Human Physiology, Mc Graw Hill
International Ed., 11th Ed., New York, USA.

8697

VOL. 10, NO 19, OCTOBER, 2015 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences

©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Jambhekar, N. J. 2011. Red Blood Cells Classification
using Image Processing. Science Research Reporter, 1(3),
pp. 151-154.

Mahmood, N.H. Poon C.L., Mazalan, S.M. and Abdul
Razak, M. A. 2013. Blood Cells Extraction using Color
Based Segmentation Technique. International Journal of
Life Science Biotechnology and Pharma Research, 2(2),
pp. 233-240.

Natsution A.M.T, and Suryaningtyas, E.K. 2008.
Comparison of Red Blood Cells Counting using two
Algorithms: Connected Component Labelling and
Backprojection of Artificial Neural Networ., IEEE,
Singapore, pp. 1-4.

Poomcokrak, J. and Neatpisarnvanit, C. 2008. Red Blood
Cells Extraction and Counting. The 3rd International
Symposium on Biomedical Engineering.

Sharif, J. M., Miswan, M. F., Ngadi, M. A., Salam, M. S.
and Abdul Jamil, M. M. 2012. Red Blood Cell
Segmentation Using Masking and Watershed Algorithm:
A Preliminary Study. International Conference of
Biomedical Engineering (ICoBE), Penang, Malaysia.

Tomari, R., Wan Zakaria, W. N., Abdul Jamil, M. M.,
Mohd Nor, F. and Nik Fuad, N. F. 2014. Computer Aided
System for Red Blood Cell Classification in Blood Smear
Image. Procedia Computer Science, 42, pp. 206-213.

Tomari R., Zakaria W.N.W, Ngadengon R. and Wahab
M.H.A. 2015. Red Blood Cell Counting Analysis by
Considering an Overlapping Constraint, ARPN Journal of
Engineering and Applied Sciences, 10(3), pp. 1413-1420.

Venkatalakshmi, B. and Thilagavathi, K. 2013. Automatic
Red Blood Cell Counting using Hough Transform. IEEE
Conference on Information and Communication
Technologies (ICT).

Vromen, J. and McCane, B. 2009. Red Blood Cell
Segmentation from SEM Images. 24th International
Conference Image and Vision Computing, New Zealand.

Wang, R., MacCane, B. and Fang, B. 2010. RBC Image
Segmentation based on Shape Reconstruction and Multi-
scale Surface Fitting. 3rd International Symposium of
Information Science and Engineering, pp. 586-589.

Wang, R. and MacCane, B. 2008. Red Blood Cell
Classification through Depth Map and Surface Feature.
International Symposium on Computer Science and
Computational Technology, pp. 339-342.

8698

2010 IEEE EMBS Conference on Biomedical Engineering & Sciences (IECBES 2010), Kuala Lumpur, Malaysia, 30th November - 2nd December 2010.

Screws Placement Effect on Locking Compression
Plate (LCP) for Tibial Oblique Fracture Fixation

lRaja Mohd Aizat Raja Izaham Darhaysham AI-Jefri Muslim
2Mohammed Rafiq Abdul Kadir Department of Orthopaedics, School of Medical Sciences,

Medical Implant Technology Group, Universiti Sains Malaysia,
Faculty of Biomedical & Health Sciences, Health Campus, Kubang Kerian, Kelantan 16150, Malaysia.
Universiti Teknologi Malaysia, 81310, Johor Bahru,
[email protected]
Malaysia.

I [email protected]
[email protected]

Abstract-One of the basic principles for internal fixation is the Compression Plate (OCP) and the Locking Compression Plate
achievement of stable construct for proper bone healing. Locking (LCP). LCP evolved through the notion that in order to obtain
compression plate (LCP) achieved it through locking screws that fast and satisfactory fracture healing, it was not always
create a fixed-angle construct providing angular stability. The necessary to achieve absolute stability through rigid internal
plate and screws combination to construct a stable fixation fixation [3].
depends on factors such as the type of fracture and surgical
preference. The optimum combination can provide optimum The unique feature of LCP is the existence of threaded
result in terms of achieving boney union. In this study, three holes that can only fit with locking head screws (LHS). Whilst
combinations of screw placement with only one screw for each conventional DCP plate requires the device to be in close
bone fragment for a fracture fixation construct using LCP were contact with the fractured bone segments, this condition,
analysed via finite element method. Three dimensional model of however, is not required for LCP [4]. Therefore the plate does
a tibia and fibula were reconstructed from computed not need to follow the contour of the bone surface which will
tomography image datasets, and an oblique fracture was decrease the chances for loss of reduction [3]. Since the plate
simulated at the midshaft of the tibia. Eight-hole LCP was placed is not compressed to the bone surface, there will be no damage
across the fractured line and three screws placement were to the periosteum and the blood supply [4]. Others have
simulated. The properties of bone were assigned with a stiffness reported that the use of LCP greatly enhances fixation
of 17GPa and Poisson's ratio of 0.3, ligaments were modelled as especially in low mineral density bone which causes
rigid links, and load equivalent to three times the body weight rheumatoid arthritis and osteoporotic bone and reduces the
was applied equally to the tibial plateau. Results showed that risk of implant failure from the screw pullout [3, 5-7].
there placement of screws the closest to the fracture side will
provide more stabiliy and less stress to the implant and the The recent LCP plate being manufacture with the
fracture construct itself. The results are unique to the simulated combination holes which can house the locking screws and
oblique fracture, and other types of fracture could also be the conventional bone screws [4]. As a result, the LCP plate
analysed using the same method. can act as the compression plate with the usage of
conventional screws or locking plate with the usage of locking
Keywords- finite element analysis (fea); locking compression screws. Therefore maximum number of combination and
plate; tibia; obliquefracture option with the usage of LCP for bone fracture fixation is
possible[8].
T. INTRODUCTION
The optimum number of screws and their placement
Human tibia has a higher tendency to fracture during depended on many factors and is still a debate for orthopaedic
accidents or falls due to its slender shape [1]. The fracture surgeons [9-11]. Some of the factors include the surgeon's
fixation method usually depends on the state of injury and the experience, fracture pattern and condition of the bone as well
type of fracture. Direct bone healing is the most common as the patient [8, 12]. This study will address the issue of
method, where the internal fixation device is applied directly optimum numbers and placement of screws for a particular
to the bone. This plate osteosynthesis device is still recognized type of tibial fracture through computational [mite element
as the treatment of choice for most articular fractures, many method. Results in terms of stress distribution and stability
metaphyseal fractures and certain diaphyseal fracture [2]. can be utilised for the advancement of fracture management
and development of new and optimised plate.
The fixation device and the surgical techniques for internal
fixation have evolved since 1960s to provide better and
improved bone healing [2]. Two major types of fixation
device that are available for fracture fixation are the Dynamic

978-1-4244-7600-8/10/$26.00 ©2010 IEEE 236

II. METHODOLOGY 10mm

A. Three Dimensional Model Design 0

Three dimensional (3D) model of human tibia and fibula 0
was reconstructed from two dimensional Computed
Tomography (CT) image dataset using 3D model 0 EEll
reconstruction software. Medullary canal for the tibia was also 0 b:
modelled along the epiphyseal segment. Oblique fracture was
simulated with an angle of 40° from lateral to medial at the 3
middle of the tibial shaft [13]. The bone model and plate
positioning are shown in figure 1. 03

0

4mm

0

0

Figure. 2 LCP and locking screw design

B. Material Properties

Two of the most common materials used for the
manufacture of LCP are stainless steel and titanium alloy. For
this particular simulation, titanium alloys (Ti-6AL-4V) was
chosen as the material for the plate and locking screws. The
material properties of the plate system and the bones were
taken from other previous studies [15-16]. AII models were
assumed to be linear elastic, isotropic and homogeneous

(Table 1).

TABLE I
MATERlAL PROPERTIES OF THE RECONSTRUCTED 3D MODEL

Materials Young's Modulus, E Poisson Ratio,
(MPa)
1)
LCP 110 000
0.3
Figure. I The position of the LCP plate on the 3D model of simulated tibial Locking screws 110 000 0.3
obIique fracture. 0.3
Tibia ___1:-= 7::-.000 0.3
Locking compression plate with 8 combination holes will
be used in the simulation of the tibia oblique fracture with Fibula 17 000
4.5mm locking screws. The design of both the plate and screw
was made using a commercial 3D modelling software (Figure. C. Analysis

2). As this study concentrated on macro scale, only the effect Bones and implants were surfaced meshed with 1.0mm
equilateral triangle. The instrumented bone model were
of screw head locking threads will be simulated. positioned in such a way that the medullary canal is aligned
LCP plate was positioned at the medial side where the with the global z-axis. The model was rigidly fixed at the
distal end of the tibia in all degrees of freedom. Tibio-fibular
bone is not in tension [14]. The plate was placed along the ligaments were modeled using rigid links to connect the fibula
tibial shaft with the middle of the plate adjacent to the with the tibia.

fractured site. The 2 holes in the middle were left empty for Magnitudes of the muscles and ligaments as well as the
joint contact forces were taken at 45% gait cycle, the point of
all analysis. second peak in ground reactions during gait [15-16]. Point
load was used to represent the magnitude of all forces as
shown in table IT.

237

TABLE II
MAGNITUDE OF MUSCLE AND JOINT CONTACT FORCES OF THE KNEE AND ANKLE [16].

Forces Forces (N)

lIIiotibial tract I x �Y _______________ Z
�---------- --- -8.5 8.8 61.3
lIIiotibial tract
II -97.4 64.4 291.5

Quadrice s femoris m. 13.6 ---�------ 32.8 303.5
Tibialis Anterior m. I -�----l;ro 7-=.2':;" : -38.7 -327.7

Tibialis Anterior m. II 25.9 -53.6 -191.8

Soleus m. 47.1�----�-------- � ---- -------

Ant. Tibiofibular lig. -63.1 -679.0

Ant. Cruciate lig -132.4 11l.2 -56.8

Deltoid Lig. ___ --0:87.5 4__-1_ 00,- '" 1.;:2�________ l"' " .1

44.9 -9.7 15.7
Knee -214-: .9� -----------------1":' 5-=-2::: -8:-: : .1
232.3
Ankle
-120.0 154.4 2070.4

All models were then meshed with 4-node structural solid IS the highest, followed by combination 2 and finally
tetrahedral elements. The bone models had a total of 207,282 combination 3 the lowest mean value.
elements whilst the LCP had 26,074 elements. All locking
screws has an average of 1,523 elements. The size of the 1400
tetrahedral mesh was made consistent for all the models [17].
Ii
The two fractured tibial segments were modelled with a
perfect contact at the start of the analysis and were allowed to a..
move relative to each other. Connection between the tibia and
fibula was through rigid links that simulate the ligaments. -:i! lZoo
Since LCP does not need to be in contact with the bone o'"n
surface, no contact was modelled between the plate and the
bone. The threads on locking screws and thread holes were 41
ignored, however, their effects of providing strong fixation
were simulated. All contacting surfaces were assigned with a ::: 1000
friction coefficient of 0.3.
<II
TABLE III
NUMBERS OF SCREWS AND THEIR LOCATIONS ON THE PLATE. '"
4'"1

� 8.00

c

� 6.00

C
..!.!
.:2 4.00

:::J
.D.'.
.C. Z.OO

41

:i! 0.00

Combl Comb2 Conb3

Figure. 3 Mean for Equivalent Von Mises Stress for case combination in the

analysis

Three combinations of screw placement were analysed in Equivalent Von Mises stress distribution of the seven
this study as shown in table TIT. The stability and effectiveness combinations shows the same pattern. The distribution origin
of each LCP-screw combination were compared from the from the middle of the plate between the 4th and 5th screw
results of Equivalent von Mises stress and displacement. hole, where the plate cover the fracture side. Even with the
same pattern of stress distribution, the magnitude of each
TIT. RESULT stress was different for all the combination (figure. 4). The
stress distribution on the plate can be see clearly on figure 4,
Higher stress was seen on the plate between the 2 screw where combination Iholds the wider stress distribution follow
holes. Besides that high stress can be seen on the screws head by combination 2 and combination 3 with small area being
especially on screws that attach with the distal part of the covered by stress.
fracture bone. Stress only can be seen on implant where else
the stress on the bone was too minimal and can be ignored. Although the stress distribution can be seen clearly on the
Combination 1 holds the highest mean stress value with 1l.43 implant, meanwhile stress distribution on the bone itself only
MPa and then combination 3 with 2.95 MPa mean stress was can be seen on the screws holes on the bones and small
the lowest stress value (figure. 3). Mean stress value of the portion of the stress distribution on the edge of fracture side.
combination can be arranged decreasingly with combination 1

238















R: How did you study in class with respect to programming…(pause).. until I found the correct
programing courses? answer.
R: Do you think that is important for you?
S2: Hmm.. In Programming courses, I have to focus, S10: Yes it is, hmm.. I like programming very much. I
concentrate and hmm.. do a lot of exercise to make feel not very happy if I not finish my job…hmm..
myself understand the topics clearly. for example ..aaa.. I didn’t finish my assignment or
project yet. Hmm.. That’s’ why I force myself to
R: Can you give reason for that? work hard on it.
S2: (Pause…) Hmm… I do that because of
Students S14 and S10 response show their approach to
programming is quite difficult course…(pause..) learning experience on these courses.
compare to other courses.
R: Did you face any problem during exercise related VII.
to programming course?
S2: Yes I do, Hmm… sometimes I get confuse which Teaching and assessment methods often encourage a
command should I used to implement for the surface approach when they are not aligned to the outcomes
specific function. of the course. As in this preliminary case study, from the
total number of students in year 1 and 2, 56% of students
This interview shows how student S2 talks about his use surface learning approach whereby only 44% applied
experience in these courses. the deep learning approach. The percentages of average
score for deep learning is greater than the average score of
This second excerpt is from an interview with a female surface learning which are 58.9% and 56.8% respectively.
first year student, S14, who is currently studying Analysis from the Chi Square Test found that there is
programming courses. insufficient evidence to conclude that there is a relationship
between Approach to learning and Year of study ( 2 =2.95, P
. value 0,157) It appears that both year 1 and year 2 apply deep
R: What did you do in class during lecture session? approach of learning.
S14: Hmm.. (pause..) I did whatever lecturer told me to
From the qualitative perspective, results from face-to-
do so. face interview with some students indicated that students
R: Anything else? found the programming courses difficult and sometimes
S14: Anything else… aaa…Writing important notes hm.. quite confusing. Most of the students said that they only did
whatever the lecturer asked them to do such as exercises and
and (pause..) hmm.. tries to understand topics practical tasks. In addition, majority of the students studied
taught by lecturer. and memorized related topics as informed by the lecturer for
R: What did you do other class times with regards to test and examination. These results showed that students
programming courses? only perform rote learning and information reproducing
S14: Aaaa….(pause) So far, I didn’t do anything. which are described as surface learning. Nevertheless, there
were students who showed positive motivation and strategy
R: How about preparation for test and examination? with an intention to understand information vigorously.
S14: For examination…hmm…I have to read and then
Throughout the quantitative and qualitative preliminary
memorize all related theories, fact, symbols and study, we can see that most of the students apply both
syntax of the programming…(pause) to answer the approaches to learning during the learning process. The
questions in examination. approaches can be switched from one approach to the other
R: Did you start from the beginning of the course, get according to various types of factors consisting of
more resource? contextual factors, perceived contextual factors and student
S14: Aaaa…. No, I did not fine other resource,..(pause..) factors [6].
just lecture notes and ..aaa… I didn’t start from the
beginning of the course; hmm… I start study and The presence of widespread and frequent use of surface
focus on whatever topics related to clue or hint learning approach signals that something is out of kilter in
given by lecturer. our teaching or in our assessment methods, but that it is
something we hope to address. The approaches that prevail
This third excerpt is from an individual interview with a tell us something about the quality of the teaching
male Second year student, S10, who has completed his environment that needs to be taken into consideration.
study on programming courses, and is chosen as it is a non- Furthermore, the lack of consistency in the findings
typical sample. highlights the need for more in-depth analysis of the effect
of learning strategies on academic performance [18].
R: What did you do when lecturer give you an
assignment or project?

S10: Firstly, I have to search for literature, then
performing simulation ..aaa…(pause).. by doing
‘try and error’ method during

104