See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/44279775

Controlling Wheelchair Using Electroencephalogram

Article  in  International Journal of Computer Science and Information Security, · May 2010

Source: DOAJ

CITATIONS READS
3 136

4 authors, including:

Vijay Khare Manvir Bhatia

Jaypee Institute of Information Technology neurology and sleep centre ,delhi
47 PUBLICATIONS   89 CITATIONS    97 PUBLICATIONS   1,390 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Analysis and Identification of Parkinson disease based on fMRI View project

Analysis and Identification of Parkinson disease based on fMRI View project

All content following this page was uploaded by Vijay Khare on 29 June 2016.

The user has requested enhancement of the downloaded file.

 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010

Controlling Wheelchair Using Electroencephalogram

Vijay Khare1 Jayashree Santhosh2

Dept. of Electronics and Communication, Engineering Computer ServicesCentre
Jaypee Institute of Information Technology Indian Institute of Technology,
Nioda, India Delhi, India
Email : vijay.khare@jiit.ac.in Email : jayashree@cc.iitd.ac.in

Sneh Anand3 Manvir Bhatia4

Centre for Biomedical Engineering Centre Department of Sleep Medicine,
Indian Institute of Technology, Sir Ganga Ram Hospital,
Delhi, India New Delhi, India
Email : sneh@iitd.ernet.in Email : manvirbhatia1@yahoo.com

Abstract— This paper present the development of a power This paper introduces the working prototype of a Brain
wheelchair controller based on Electroencephalogram (EEG).To Controlled Wheelchair (BCW) that can navigate inside a
achieve this goal wavelet packet transform (WPT) was used for typical office and hospital environment with minimum
feature extraction of the relevant frequency bands from structural modification. It is safe and relatively low cost and
electroencephalogram (EEG) signals. Radial Basis Function provides optimal interaction between the user and wheelchair
network was used to classify the pre defined movements such as within the constraints of brain computer interface.
rest, forward, backward, left and right of the wheelchair.
In this study, Wavelet Packet Transform (WPT) method
Classification and evaluation results showed the feasibility of
EEG as an input interface to control a mechanical device like
was used for feature extraction of mental tasks from eight
powered wheelchair. channel EEG signals. WPT coefficients give the best
discrimination between the directions of wheelchair in the
Keywords— Electroencephalogram (EEG), Wavelet Packet relevant frequency band. The WPT coefficients were used as
Transform (WPT), Radial Basis Function neural network (RBFNN), the best fitting input vector for classifier. Radial Basis
Brain computer interface (BCI), Rehabilitation, Wheelchair Function network was used to classify the signals.
Controller.

II. METHODOLOGY
I. INTRODUCTION
There are numerous interfaces and communication A. Subjects
methods between human and machines. A typical human Nine right-handed healthy male subjects of age
interface utilizes input devices such as keyboard, mouse, (mean: 23yr) having no sign of any motor- neuron diseases
joystick, chin control, ultrasonic non contact head controller were selected for the study. A pro-forma was filled in with
and voice controller. Such interfaces were developed to detail of their age & education level as shown in Table I. The
improve manipulability, safety and comfortness. Literature participants were student volunteers for their availability and
survey shows existing systems such as Chin controller is interest in the study. EEG data was collected after taking
inconvenient to use, ultrasonic non-contact head controller has written consent for participation. Full explanation of the
relatively low accuracy and voice controller gives delayed experiment was provided to each of the participants.
response to voice command hence not useful in noisy
environment[1-2]. Recently, a number of biological signals
such as electromyogram (EMG), Electroencephalogram (EEG)
and Electrocculogram (E.O.G) have been employed as hands- TABLE I. CLINICAL CHARACTERISTICS OF SUBJECTS
free interface to machines [3-7]. Brain Computer Interface
(BCI) system has been shown to have the potential to offer S.No. Subject Age Educational
status
humans a new nonmuscular communication channel, which
enables the user to communicate with their external 1 Subject 1 22 BE
surroundings using the brain’s electrical activity measured as
electroencephalogram (EEG) [8-12]. 2 Subject 2 21 BE

181 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010
3 Subject 3 23 BE minute (as shown in Fig 2). The whole experiment lasted for
about one hour including electrode placement.
4 Subject 4 27 M.TECH

5 Subject 5 23 BE

6 Subject 6 22 BE

7 Subject 7 27 M.TECH 10ms

10ms

8 Subject 8 22 BE
Sec……..
0 10 20
Mental
9 Subject 9 22 BE Relax Task

B. EEG Data Acquisition Figure 2: Timing of the Protocol

EEG Data used in this study was recorded on a Data collected from nine subjects performing five mental
Grass Telefactor EEG Twin3 Machine available at Deptt. of tasks were analyzed. The following mental tasks were used to
Neurology , Sir Ganga Ram Hospital, New Delhi. EEG record the appropriate EEG data.
recording for nine selected subjects were done for five mental
tasks for five days. Data was recorded for 10 sec during each
• Movement Imagination:-The subject was asked to
task and each task was repeated five times per session per day.
Bipolar and Referential EEG was recorded using eight plan movement of the right hand.
standard positions C3, C4, P3, P4, O1 O2, and F3, F4 by
placing gold electrodes on scalp, as per the international • Geometric Figure Rotation:-The subject was given 30
standard 10-20 system of electrode placement as shown in Fig seconds to see a complex three dimensional object,
1. The reference electrodes were placed on ear lobes and after which the object was removed. The subject was
ground electrode on forehead. EOG (Electooculargram) being instructed to visualize the object being rotated about
a noise artifact, was derived from two electrodes placed on an axis.
outer canthus of left and right eye in order to detect and
eliminate eye movement artifact. The settings used for data • Arithmetic Task:-The subject was asked to perform
collection were: low pass filter 1Hz, high pass filter 35 Hz, trivial and nontrivial multiplication. An example of a
sensitivity 150 micro volts/mm and sampling frequency fixed trivial calculation is to multiply 2 by 3 and nontrivial
at 400 Hz. task is to multiply 49 by 78. The subject was
instructed not to vocalize or make movements while
solving the problem.
• Relaxed: - The subject was asked to relax with eyes
F3 F4 closed. No mental or physical task to be performed at
this stage.
C3 C4
D. Feature Extraction
A1 A2
P3 P4 The frequency spectrum of the signal was first
analyzed through Fast Fourier Transform (FFT) method [13-
O1 O2 14]. The FFT plots of signals from all the electrode pairs were
observed and maximum average change in EEG amplitude
was noted as shown in Fig3. For relaxed state, the peak of
Figure1:- Montage for present study power spectrum almost coincides at for central and occipital
C. Experiment Paradigm area in the alpha frequency range (8-13Hz) [15]. EEG
recorded with relaxed state is considered to be the base line for
An experiment paradigm was designed for the study the subsequent analysis. Mu rhythms are generated over
and the protocol was explained to each participant before sensorimotor cortex during planning a movement. For
conducting the experiment. In this, the subject was asked to movement imagery of right hand, maximum upto 50% band
comfortably lie down in a relaxed position with eyes closed. power attenuation was observed in contralateral (C3 w.r.t C4)
After assuring the normal relaxed state by checking the status hemisphere in the alpha frequency range (8-13Hz) [16]. For
of alpha waves, the EEG was recorded for 50 sec, collecting geometrical figure rotation, the peak of the power spectrum
five session of 10sec epoch each for the relaxed state. This was increased in right hemisphere rather than left in the
was used as the baseline reference for further analysis of occipital area for the alpha frequency range (8-13Hz)[17]. For
mental task. The subject was asked to perform a mental task trivial multiplication, the peak of the power spectrum was
on presentation of an audio cue. Five session of 10sec epoch increased in left hemisphere rather than right hemisphere in
for each mental task were recorded, each with a time gap of 5 the frontal area for the alpha frequency range (8-
13Hz)[18].For non trivial multiplication, the peak of the power
spectrum was increased in left hemisphere rather than right

182 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010
hemisphere in the parietal area for the alpha frequency range mapped into 3 bit as shown in table3 to provide parallel port
(8-13Hz). input bit, which was used to drive the motor

F. Hardware implementation

60
O1-O2
50
C3-C4 F3-F4 C3-C4
Amplitute (db)

F3-F4
40

30

20

10

0
Movement Trivial Rotation Nontrivial Relax
Mental Tasks
Figure 4: Conceptual block diagram of the wheelchair controlled by EEG
signals (courtesy http//www. getch.at/g.bcisys/bci.html)

Conceptual block diagram of EEG based power wheelchair

Figure 3: Maximum Average change in Amplitude of PSD
system is shown in Fig 4. Using parallel port, Motor driver IC
The data was preprocessed using Wavelet packet transform (IC L293) was interfaced with computer as shown in Fig 5 for
to extract the most relevant information from the EEG signal. the wheelchair controller. In the circuit, P1 acts to enable the
[19-20]. By applying Wavelet packet transform on the original chip and combination of P2 and P3 were used to control
signal wavelet coefficients in the (8-13Hz) frequency band at direction of wheelchair. The truth table for the above logic is
the 5th level node (5, 3) were obtained. Twenty one shown in Table III with polarities of motor of M1and M2. All
coefficients have been obtained from one second of EEG data. five direction of wheelchair movement were properly
These coefficients are scaled and used as the best fitting input controlled by this designed circuit.
vector for classifiers. Subsequently the signal was
reconstructed at node (5, 3).

E. Classifier
For classification, Radial Basis Function Neural
Network (RBFNN) classifier was employed. A two layer
network was implemented with 21 input vectors, a hidden layer
with Gaussian activation function consisting as many as hidden
neurons as input vectors and five neuron in the output layer
[21-23]. RBFNN produces a network with zero error on
training vectors. Using RBFNN the five mental tasks were
classified, as shown in Tables II

Figure 5: Circuit Diagram for wheelchair controller

TABLE II. CLASSIFICATION OF FOUR MENTAL TASKS
Tasks Accuracy classifications
TABLE III. TRUTH TABLE OF HARDWARE DESIGN
%
Movement Imagery 100 00100 P2 P3 M1 M2 TASKS
P1
Trivial Multiplication
Geometric Figure
100
100
01000
00010
+ - + -
1 0 0 0 1 1 0 LEFT (L)
Rotation
Nontrivial Multiplication 100 10000
1 1 0 1 0 1 0 FORWARD(F)
Relax 100 00001
1 0 1 0 1 0 1 BACKWARD(B)

After the classification of five mental tasks namely 1 1 1 1 0 0 1 RIGHT (R)

movement imagery, trivial multiplication, geometrical figure
rotation, nontrivial multiplication and relax, the output of the
0 X X --- --- --- --- STOP(S)
classifier was interfaced with the motor using parallel port. The
motor driver required 3 bit of data. The output of classifier was

183 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010
• For nontrivial multiplication task, the amplitude of
the power spectrum increases in the left parietal
region for alpha frequency range (8-13Hz).
This observation could be used successfully for controlling
power wheelchair. It can be noted that comparing Table
IV&V, there is perfect match with earlier studies and changes
are prominent and unique.

Figure 6: State diagram for Wheelchair Movement

The polarities of the motors M1 and M2 are shown in the

truth table. State diagram for wheel chair movement in
different direction is shown in Fig 6. For movement imagery
task, the output of parallel port would be [1 0 0]. Due to
opposite polarities, M2 motor would move forward and
M1motor backward which would be lead to left movement of
the wheelchair. For trivial multiplication task, the output of
parallel port would be [1 1 0]. Due to same polarities, both Figure 7: Movement task classification
motors M1and M2 move forward resulting forward movement
of the wheelchair. For geometrical figure rotation task, the
output of parallel port would be [1 1 1]. Due to opposite
polarities, M1 motor moves forward and M2 motor backward
resulting right movement of the wheelchair. For nontrivial
multiplication, the output of parallel port would be [1 0 1].
Due to same polarities, both motors M1and M2 move
backward resulting backward movement of the wheelchair.
Similarly, for stop tasks output of parallel port would be [0 x
x] and the wheelchair would be control by different polarities
at the motors.

III. RESULT AND DISCUSSION

Classification of five mental tasks shown in the Fig (7-11)
Earlier researchers had established [15-18] the most prominent Figure8: Trivial multiplication task classification
areas in brain for domain of information during various mental
tasks as shown in Table IV. In the present study, maximum
average change in EEG amplitude has been observed by us as
shown in TableV. The study had led to following
observations:

• For movement imagery task, the amplitude of the

power spectrum for alpha frequency range (8-13Hz)
had attenuation in contralateral area.

• For geometrical figure rotation task, the amplitude of

the power spectrum increases in the right occipital
region for alpha frequency range (8-13Hz).

• For trivial multiplication task, the amplitude of the Figure9: Geometrical figure rotation classification
power spectrum increases in the left frontal region for
alpha frequency range (8-13Hz).

184 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010
TABLE V. AVERAGED AMPLITUDE IN ALPHA FREQUENCY FOR NINE
SUBJECT

Tasks Electrodes Amplitude(db) in alpha

rhythm(8-13Hz)

Movement
Imagination C3 C4 C3(33.75) C4(43.02)

Trivial
Multiplication F3 F4 F3(43.34) F4(39.69)
Geometrical figure
rotational O1 O2 O1(42.35) O2(52.43)

Figure10: Non trivial task classification

Non-trivial
Multiplication P3 P4 P3(40.03) P4(34.74)

Relax C3 C4 C3(42.29) C4(44.01)

Fig 12(a-d) associated with table 6 show four experiments

conducted on nine right-handed male subjects. The subjects
were asked to mentally drive the wheelchair from the starting
point to a goal by executing the five different mental tasks
namely Movement Imagery (MI), Trivial Multiplication(TM),
Geometrical Figure Rotation (GFR), Non Trivial
Multiplication (NTM) and Relax (R) to control direction of the
power wheelchair.
Figure11 Relax task classification To complete task from staring point to goal, the subject
performed sequence of the mental tasks as shown in TableVI.
Experiment has been successfully completed with 100%
TABLE IV. DOMAIN OF INFORMATION accuracy by all nine subjects.
Tasks Domain of (Contralateral/ Type of change in
information Ipsilateral) amplitude of
TABLE VI. MATRIX OF MENTAL TASKS AND DIRECTION OF WHEELCHAIR
alpha rhythm(8-
13Hz)
Movement Central Contralateral Decreased Path a TM/ GFR/ GFR/Left GFR/ R/ Stop
Imagination Forward Left Left

Arithmetic Frontal, Ipsilateral Increased Path b TM/ MI/ MI/ MI/ R/ Stop
Simple Forward Right Right Right

Geometrical Occipital Ipsilateral Increased Path c TM/ GFR/ GFR/Left GFR/ R/ Stop
figure rotational Forward Left
Left
Arithmetic parietal Ipsilateral Increased
complex Path d TM/ MI/ GFR/Left GFR/ R/ Stop
Forward Right Left
Base line Occipital, Contralateral Coincide
Central

185 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010
[3] F. Galan, M. Nuttan, E. Lew, P.W. Ferrez, G. Vanacker, J. Philip, J. del
R.Millan, “A brain actuated wheelchair: Asynchronous and nojn
invasive brain computer interfaces for continuous control of robot”
,Clinical Neurophysiology, Vol 119, pp. 2159-2169, 2008.
[4] A.T.C. Au and R.F. Kirsch, “EMG based prediction of shoulder and
elbow kinematics in able –bodies and spinal cord injured individual,”
IEEE Trans. Rehab.Eng.vol.8 no.4, pp. 471-480, Dec.2000.
Start Goal [5] J. Millan, “Noninvasive brain actuated control of a mobile robot by
Goal Star
Goal human EEG,” IEEE Trans. Biomed. Eng, vol. 51, no 6, pp.1026-1033,
June 2004.
a b
[6] R. Bare, “E.O.G guidance of a wheeelchair using neural network,” in
porc.Int.Conf. Pattern recognition, Barcelona Spain, pp. 4668-4672,
2000.
[7] R. Leeb, “Self paced (Asynchronous) BCI-Contol of a wheelchair in
virtual environment :Acase study with tetraplegic”, Computational
Intelligent & Neuroscience.Vol 2007, Aritical ID 79642, 2007
[8] F. Lotte, M. Congedo, A. Lecuyer, F. Lamarche, B Arnaldi, “A Review
of Classification algorithms for EEG bases brain computer interface,”
Journal of neural Engineering, Vol. 4, R1-R13, 2007.
Goal
Goal
Start [9] R. Boostani, B. Graimann, M.H. Moradi, G. Plurfscheller, “ Comparison
Start approach toward finding the best feature and classifier in cue BCI”,
c d
Med. Bio. Engg., Computer Vol 45, pp. 403-413, 2007.
[10] J.R. Wolpaw, N. Birbaumer, D.J. Mc Farland, G.Plurtscheller, T.M.
Vaughan, “Brain computer Interfaces for communication and
Figure12(a-d): Top view of random path control”, Clinical Neurophys. , Vol 113, pp. 767-791, 2002.
[11] G. Pfurtschelle, D. Flotzinger, and J.Kalcher, “Brain Computer
interface-A new communication device for handicapped people,” J
Microcomput. Applicate., vol. 16, pp. 293-299, 1993.
IV. CONCLUSION [12] J. R. Wolpaw, T. M. Vaughan and E. Donchin, “EEG Based
Communication prospects and problems,” IEEE, Trans.Rehab. Eng.,
The present study investigated controlling of a power Vol.4, pp. 425-430, Dec.1996.
wheelchair by EEG signals. This was an attempt to control [13] Z. A. Keirn and J. I. Aunon, “A new mode of communication between
direction of wheelchair via brain signals. Each direction (left, man and his surroundings,” IEEE Trans. Biomed. Eng., Vol. 37, No. 12,
forward, right, backward and stop) of the wheelchair pp. 1209–1214, Dec. 1990.
corresponded to five mental tasks (movement imagery, trivial [14] R. Palaniappan, “Brain computer interface design using band powers
multiplication, geometrical figure rotation, non -trivial extracted during mental task,” proceeding of the 2ndInternational IEEE
EMBS Conference on Neural Engineering , pp. 321- 324, 2005.
multiplication and relax). To differentiate five mental tasks,
[15] G. Pfurtscheller, C. Neuper, A. Schlogl and K. Lugger, “Separability of
wavelet packet transform was employed for feature extraction EEG signals recorded during right and left motor imagery using adaptive
and Radial basis function neural network was used for auto regressive parameters, IEEE. Trans. on rehabilitation
classification. The experimental result showed 100% accuracy. Engineering, Vol 6 ,No3, pp. 316-325, 1998.
[16] J. Santhosh, M. Bhatia, S. Sahu, S. Anand, “Quantitative EEG analysis
for assessment to plan a task in ALS patients, a study of executive
The authors would like to extend the work with severely function (planning) in ALS,” Cognitive brain research Vol 22, pp. 59-
disabled people and to customize the device as per individual 66, 2004.
response and requirements. This kind of system can also be [17] A. R. Nikolaev and A. P. Anokhin, “EEG frequency ranges during
used in a variety of applications like– reception and mental rotation of two and three dimensional objects,”
Neuroscience and Bheaviour physiology, Vol. 28, No-6, 1998.
¾ Environment control units (ECU’S)
[18] Osaka M., “Peak alpha frequency of EEG during a mental task: task
¾ Helping disable people to directly interact with hand difficulty and hemisphere difference,” Psychophysiology, Vol.21, pp.
held devices such as cell phones and PDAs. 101-105, 1984.
¾ Dealing with hazardous material/chemical at [19] C. S. Li and H. Wang, “Wavelet transform for on –off switching BCI
laboratories. device,” 7th Asian-Pacific Conference on Medical and Biological
Engineering, Beijing, China, Vol 19, pp. 363-365, 22–25 April 2008.
ACKNOWLEDGMENT [20] B. Guo Xu & A. G. Song , “Pattern recognition of motor imagery EEG
using wavelet transform,” Jouranl of Biomedical Science &Engineering,
The authors would like to acknowledge their gratitude to Vol 1, pp. 64-67, 2008.
the staff of EEG Laboratory at Sir Ganga Ram hospital, New [21] E. Larsson, K. Åhlander and A. Hall, “Multi-dimensional option pricing
Delhi for the help in carrying out the experiment. using radial basis functions and the generalized Fourier transform,” In J.
Comput. Appl. Math., 2008.
REFERENCES [22] U. Pettersson, E. Larsson, G. Marcusson and J. Persson, “Improved
radial basis function methods for multi-dimensional option,” In J.
[1] K. Tanaka. K. Matsunaga, N. Kanamori, S. Hori, and H.O.Wang, Comput. Appl. Math., 2008.
“Electroencephalogram based control of a mobile robot,” in proc. IEEE [23] S. Chen, C. F. N. Cowan and P. M. Grant, "Orthogonal Least Squares
Int. Symp. Computational Intell.. Robot.Autom Kobe, Japan, pp 670- Learning Algorithm for Radial Basis Function Networks," IEEE
675, Jul.2003. Transactions on Neural Networks, vol. 2, No. 2, pp. 302-309, March
[2] K. choi & A. Cichocki , “Control a Wheelchair by Motor Imagery in 1991.
Real Time,” IDEAL 2008, LNCS 5326 , pp. 330-337, Springer Verlag
Berlin Heidelberg 2008.

186 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No.2, 2010
AUTHORS PROFILE Prof. Sneh Anand is a professor and head, Center
for Biomedical Engineering, Indian Institute of
Vijay Khare is currently pursuing his PhD in Bio Technology, Delhi. She did B.Tech in Electrical Engg,
Signal Processing at the Indian Institute of Technology, from Punjab University, Patiala, and M.Tech in
Delhi. He did his M.Tech in Instrumentation & Control, Instrumentation & Control from IIT Delhi and Ph.D. in
from NSIT Delhi. He is currently, with the Dept. Biomedical Engg. from IIT Delhi. Her research interests
Electronics and Communications Engineering at the include biomedical instrumentation, rehabilitation
Jaypee Institute of Information Technology. His engineering, biomedical transducers and Sensors.
research interests are Neural Networks, Brain Computer
Interfacing, and Control Systems.

Dr.Jayashree Santhosh completed her B.Tech in

Electrical Engineering from University of Kerala, M Dr. Manvir Bhatia is the Chairperson of Dept. of
Tech in Computer & Information Sciences from Cochin SleepMedicine at Sir Ganga Ram Hospital, New Delhi
University of Science and Technology, Kerala and Ph.D and is also a Senior Consultant Neurologist.Dr. Manvir
from IIT Delhi. She is a Fellow member of IETE, Life Bhatia completed her MBBS in 1981, and Doctor of
member of Indian Association of Medical Informatics Medicine in 1986 from Christian Medical College and
(IAMI) and Indian Society of Biomechanics (ISB). Her research interests Hospital, Ludhiana. DM in Neurology 1993, from All
include IT in Healthcare Systems and was associated with a project on IT in India Institute of Medical Sciences.She is a member of
Health Care at City University of Hong Kong. She is also associated with Indian Academy of Neurology, Indian Epilepsy Society, Indian Sleep
various projects with Centre for Bio-Medical Engineering at IIT Delhi in the Disorders Association, World Association of Sleep Medicine, International
area of Technology in Healthcare. H e r r e s e a r c h i n t e r e s t s focus on Brain Restless Legs Society Study Group and American Academy of
Computer Interface Systems for the Handicapped and in Neuroscience. Electrodiagnostic Medicine.Dr. Manvir Bhatia has been invited to deliver
lectures in National & International workshops, conferences on topics related
to Neurology, Epilepsy, Sleep Medicine and has sleep published papers in
leading journals.

187 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
View publication stats