Brain computer interface based smart keyboard using neurosky mindwave headset

TELKOMNIKA Telecommunication, Computing, Electronics and Control
Vol. 18, No. 2, April 2020, pp. 919~927
ISSN: 1693-6930, accredited First Grade by Kemenristekdikti, Decree No: 21/E/KPT/2018
DOI: 10.12928/TELKOMNIKA.v18i2.13993  919
Journal homepage: http://journal.uad.ac.id/index.php/TELKOMNIKA
Brain computer interface based smart keyboard
using neurosky mindwave headset
Thair A. Salih, Yasir M. Abdal
Technical Engineering College/Mosul, Northern Technical University, Iraq
Article Info ABSTRACT
Article history:
Received Aug 28, 2019
Revised Jan 10, 2020
Accepted Feb 20, 2020
In the last decade, numerous researches in the field of electro-encephalo-
graphy (EEG) and brain-computer-interface (BCI) have been accomplished.
BCI has been developed to aid disabled/partially disabled people
to efficiently communicate with the community. This paper presents
a control tool using the Neurosky Mindwave headset, which detects
brainwaves (voluntary blinks and attention) to form a brain-
computer interface (BCI) by receiving the system signals from the frontal
lobe. This paper proposed an alternative computer input device for those
disabled people (who are physically challenged) rather than the conventional
one. The work suggested to use two virtual keyboard designs. The conducted
experiment revealed a significant result in developing user printing skills on
PCs. Encouraging results (1.55-1.8 word per minute (WPM)) were obtained
in this research in comparison to other studies.
Keywords:
BCI
EEG
Neurosignal
Virtual keyboard
This is an open access article under the CC BY-SA license.
Corresponding Author:
Yasir M. Abdal,
Technical Engineering College/Department of Computer Engineering,
Northern Technical University,
Technical Complex/Al-Mounsa Street.
Email: yasir.m.abdal@ntu.edu.iq
1. INTRODUCTION
The brain is an essential part of the human body, which controls movement, behaviour and regulates
the equilibrium of the human body. The brain provides cognition; exudes emotions; stores memories and
directs muscular or motor activities [1]. There are over 86 billion neurons in the average human brain, and
there are various other cells, which almost equal that number too. The interconnection of neurons is crucial
for brain activity, as these neurons form an associative link between brain cells [2]. Figure 1 [3], shows
the classification of the human brain, which defines regions within the brain as the forebrain, midbrain, and
hindbrain. The Midbrain mostly comprises of a portion of the brainstem, which controls some reflex actions
and is a portion of the circuit; concerned with the regulation of eye movements and other voluntary
movements. The general definition of the (brain) hemisphere happens into four lobes, namely–the frontal
lobe, parietal lobe, temporal lobe and occipital lobe [4]. Perception involves sensing the signals from
the external environment and is the major function of the human brain, which is at the core of understanding
human senses, feeling and emotion. Furthermore, regulating and controlling the human behaviours;
regulating and controlling physical actions; regulating memory functions; the method of thinking and other
reasoning processes [5]. Medical professionals use Electroencephalography (EEG) to diagnose abnormalities
in human life [6]. EEG is an electrophysiological monitoring technique that logs the electrical signals
emanating from the brain. EEG senses the electrical potential difference resulting from neurological ionic
currents produced by the brain [7]. EEG provides an automatic real-time recording of the brain's electrical

 ISSN: 1693-6930
TELKOMNIKA Telecommun Comput El Control, Vol. 18, No. 2, April 2020: 919 - 927
920
activity using multiple electrodes positioned on the scalp [8]. The past few years have seen a rapid progress
of EEG-based BCI; that meteoric rise partly attributes to computer processing and speed improvements; also,
that is due to signal analysis technique refinements. That work has produced results with applicable clinical
applications [9]. BCI is a revolutionary technology, which reads the user's mind directly from the brain and
transforms to the commands of a controllable device, bypassing the peripheral neural system [10]. The motor
imagery based BCI system is an essential type of BCI. Traditional motor images produced by BCI acquires
data through specific processes, then performs data analysis offline to choose the best channel and
parameters [11]. Invasive and Non-Invasive are two essential types of BCI arrangements. Invasive BCI
arrangements install a chip inside the brain, which records the brain activity [12]; this is often impractical
because this requires brain surgery, while non-invasive types use a headset device and are externally placed
on the scalp which measures brain activity [3].
Figure 1. The brain hemisphere
A previous study presented by [13], develops a BCI system using a Mindwave headset. That study
senses the activity of the brain and encodes the response in MATLAB, and the decision making, which
enables smart home device control uses an Arduino module. The study conducted by [14] employs a P300
BCI system to output digital text. That system displays pulsating characters and a classifier, which
determines target characters. Typically, a user must type each symbol within a word at a time. That spelling
process is slow, and it can take many minutes to output an entire word. Another research by [15], employs
a BCI system to control four-wheel electric vehicles. Emotiv EPOC+ is used to acquire the raw EEG
data.Independent Component Analysis (ICA) is used to pre-process the motor imagery EEG; feature
extraction involves a Common Spatial Pattern (CSP), which are most related to the ERD/ERS. That research
leads to the development of a non-invasive BCI, which realizes the intent of the visual cortex by equating
EEG-SSVEP signals to control a wheelchair [16] automatically. That system uses offline data analysis to
enable user motive control of the electric wheelchair.
The research conducted by [17] develops a BCI based keyboard with long-term potential for areas
such as learning, training, brain stimulation and other clinical purposes. That study presents an average
accuracy improvement of all the users from a 40% error rate in the first round to a 10.5% error rate in
the fourth round of the trials. This paper describes the use of a low-cost, non-invasive BCI type system,
which uses a Neurosky Mindwave headset. This system collects EEG data in real time, to enable people with
physical disabilities to communicate with others.
2. RESEARCH METHOD
This work successfully develops a BCI solution which detects and process brain signals in
a real-time. Five sub-blocks can represent the system, and the following is a description of
the microarchitecture design of each module shown in Figure 2.

TELKOMNIKA Telecommun Comput El Control 
Brain computer interface based smart keyboard using neurosky mindwave headset (Thair A. Salih)
921
2.1. EEG signal acquisition
The primary data acquisition element of the solution is the NeuroSkyMindWave headset. It is
an inexpensive, lightweight, portable device with wireless communication. It consists of eight parts, which
are a power switch, an ear clip, the ear arm, battery area, adjustable headband, sensor tip, sensor arm, and
think gear chip [18]. Two sensors are used to operate this device to obtain and filter EEG signals. The sensor
tip locates on the forehead and detects electrical signals emanating from the brain’s frontal lobe [19, 20].
The second sensor is an ear clip, which is used as a ground acting as a filter for electrical noise [21].
The NeuroSkyMindwave also has good measurement accuracy, which can result in a broader group of
potential users. Figure 3, explains the locations of the electrodes of the MindWave EEG headset in
the international 10/20 system [22].
Figure 2. The BCI system
Figure 3. NeuroSkyMindWave system
2.2. Preprocessing
EEG raw signals are very low power signals collected from the user scalp, amplified, digitized and
transmitted through a Bluetooth module to the personal computer using the NeuroSkyMindWave device.
2.3. Features of EEG signals
An EEG signal comprises of rhythmic activity and transients. The rhythmic activity is divided into
wave bands by frequency while the transient is referring to spontaneous spikes and wave formations that are
sharp [23]. There are five types at the most critical frequency ranges: delta (2-4Hz), theta (4-8Hz),
alpha (8-12Hz), beta (15-30Hz), and gamma (30-80Hz) [6, 24].

 ISSN: 1693-6930
922
2.4. Classification
The Neurosky uses the ThinkGear Technology to process and classify the EEG output considers
quantitative approaches to send signals via Bluetooth to the PC. The EEG processing protocols are closed
source software. The output data are Raw EEG signal, eSense Attention and Mediation is an integer value
between 0 and 100, Blink Strength is returning an integer value between 0 and 255 [18]. The "eSense
Attention" value is used to scan and initialize the virtual keyboard enabling user proper character choice.
The Blink Strength value is used to select characters or enables the cursor to move to the next row of
the virtual keyboard.
2.5. The virtual keyboard: design and work
The processing development environment (PDE) does process the data. A MindWaveWireless USB
Adapter receives the packets of data transmitted from the brain wave sensor. The Neurosky EEG sensor
obtains the attention and blink levels. If the attention value exceeds the threshold for a specific time, for
example 1 second, then the keyboard scan is 1 second, so the keyboard scan starts. The attention level is
received as a series of inputs, at a 1 Hz frequency, while the NeuroSky sensor obtains the unprocessed EEG
data at a 512 Hz frequency. The blinking level is used to select characters and enable the pointer to navigate
through the keyboard rows to hasten the writing process. The virtual keyboard is designed to contain
alphanumeric and control characters. It includes clear, space, screen and delete function characters, as shown
in Figure 4. There are thirty cells to provide a virtual keyboard, each alphabetic character and control running
a single cell. The cells are organized in a column fashion and arranged in a QWERTY design.
Figure 4. Virtual keyboard design
The red border box represents the cursor that moves on the keyboard so the user can visually
recognize the character select. Using a horizontal cursor movement, the user can select the correct character
and record within the text box in a second. Figure 5, describes a flow chart of the proposed work. Initially,
the NeuroSky headset is turned on, which in order to identify the neuro-signals. The EEG Biosensor capture
the signals and send them to the Think-Gear chip for processing. After analysis, the Java environment
receives digital signals for attentional extraction and utilizes the Blink Strength signals for further
classification. The text process uses eye blink and eye focus levels. The keyboard indicator begins to scan if
the signal level of attention exceeds the threshold for a specified period. If the Blink Strength value is
between the decimal numbers of 110 and 60, then that action chooses the text box character. If the Blink
Strength value exceeds the 110-decimal value, the pointer jumps between keyboard rows. The proposed
virtual keyboard uses one blink signal to select any row to increase text writing speed and obtain more text in
a fixed time.

923
Figure 5. Brain-controlled keyboard flow chart
3. RESULTS AND ANALYSIS
The performance of the system was evaluated by taking five people to the test, ranging in age from
30 to 35 years. During the practical test phase, each participant sits in a comfortable chair in front of
the laptop. Each participant performs the following experiments in a certain number of sessions at
different times.
3.1. First experiment
Each person was asked to write (Help) word for nine sessions on (QWERTY) virtual keyboard
designed as previously mentioned. This study calculates the time required to write the required word and
the number of wrong characters; shown in Table 1.
Table 1. Time consumed to write (HELP) word and a character error number
Sessions
Subject 1 Subject 2 Subject 3 Subject 4 Subject 5
Time
(sec)
Error
(ch)
Time
(sec)
Error
(ch)
Time
(sec)
Error
(ch)
Time
(sec)
Error
(ch)
Time
(sec)
Error
(ch)
1 136.07 5 123.06 5 120.2 4 106.5 5 83.9 4
2 129.16 4 91.95 3 99 2 92 5 70 2
3 76.1 2 80 2 69.81 2 99.11 4 54.3 0
4 58.05 1 116.3 2 69.56 2 82.05 2 72.8 4
5 47.36 0 63 1 65.5 1 77.49 1 71 3
6 50.2 1 53 1 61.2 1 72.04 1 79 3
7 47.64 0 48.2 0 60.5 1 57.5 1 44.04 0
8 46.26 0 49.3 0 55 0 48 0 53.3 1
9 45.29 0 46.2 0 56.4 0 46.9 0 44 0
The words per minute (WPM) variable provides the most frequent empirical metric of text entry
performance. It measures the period to produce a certain number of words. WPM is inconsiderate of
the number of keystrokes nor gestures types made during the text entry, but only considers text transcription
length, which is defined in (1) [25].

 ISSN: 1693-6930
924
WPM= ((T-1)/s)*60*1/5 (1)
Where S is the time in seconds. 60 is a constant. The factor of one-fifth accounts for the average word length
in characters, including spaces, numbers, and other printable characters. Table 2, shows the text entry speed
and error rate of each subject for nine sessions. Figure 6 Shows the results for every participant, during nine
sessions and calculates the average time for all sessions. Figure 7 Shows the average entry speed of text
represents as (WPM) with an accuracy of the correct entry text.
Table 2. Word entry speed and error rate
Sessions
Subject 1 Subject 2 Subject 3 Subject 4 Subject 5
Accuracy
(%)
Entry
Speed
(wpm)
Accuracy
(%)
Entry
Speed
(wpm)
Accuracy
(%)
Entry
Speed
(wpm)
Accuracy
(%)
Entry
Speed
(wpm)
Accuracy
(%)
Entry
speed
(wpm)
1 44.44 0.26 44.44 0.29 50.00 0.30 44.44 0.34 50 0.43
2 50 0.28 57.14 0.39 66.67 0.36 44.44 0.39 66.67 0.51
3 66.67 0.47 66.67 0.45 66.67 0.52 50 0.36 100 0.66
4 80 0.62 66.67 0.31 66.67 0.52 66.67 0.44 50 0.49
5 100 0.76 80 0.57 80 0.55 80 0.46 57.14 0.51
6 80 0.72 80 0.68 80 0.59 80 0.50 57.14 0.46
7 100 0.76 100 0.75 80 0.60 80 0.63 100 0.82
8 100 0.78 100 0.73 100 0.65 100 0.75 80 0.68
9 100 0.79 100 0.78 100 0.64 100 0.77 100 0.82
Figure 6. The average time for all sessions
Figure 7. Text entry speed and accuracy
0
10
20
30
40
50
60
70
80
subject 1 subject 2 subject 3 subject 4 subject 5
time(sec)

925
3.2. Second experiment
This experiment uses two types of virtual keyboards, QWERTY type as described previously, ABC
type consists of alphabets and two control characters defined as Delete and Space. The keyboard has 30 cells;
each alphabet occupies one cell character and two cells for each control button [17]. In this study, two types
of keyboard use square pointers. Both keyboards scan with a 600-millisecond delay to allow the user to select
a proper character. The persons are required to write a 24-letter sentence, including the white space character
within six sessions. The time required to write the chosen word was calculated as shown in Tables 3 and 4.
Figures 8 and 9 shows the average text entry speed and error rate of each person for this study.
The calculations of the text entry rate of various input methods is straight forward and simple.
In the first study, the users training to write the word HELP, and calculate the time taken, as shown
in Table 1. 75.83% accuracy obtained, and the WPM is 0.56, which equivalent to 2.8 letters per minute.
In the second study, the user is trained to write a sentence of 24 characters; by reducing the time taken to scan
the keyboard and switch between the buttons to be 600 milliseconds, using two types of keyboard design to
improved the proposed system. The first keyboard (QWERTY) contains three control buttons in addition to
the alphanumeric. The error rate was 5% and WPM = 1.55. The second type of keyboard designed as (ABC),
which contains two control buttons in addition to the alphanumeric. The error rate was 5.25% and
WPM = 1.8. The results of the proposed system exceed the performance of previous studies that enable
the user to write 7.75 to 9 characters per minute. Through the results obtained, it is possible to note
the improvement in the performance of the proposed system compared to the previous systems by looking at
the time taken to write most characters, taking into account the delay in the movement of the indicator
between pressing the buttons.
Table 3. Time and error data of (QWERTY) keyboard
Subjects
Session1 Session2 Session3 Session4 Session5 Session6
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
1 219.0 4 199.2 1 160.2 1 172.8 1 165.6 0 156.0 0
2 208.6 3 190.4 2 175.0 1 177.0 1 160.0 0 149.4 0
3 225.3 5 220.2 4 195.9 2 180.5 1 185.6 0 171.0 1
4 199.6 1 182.8 2 150.0 1 190.5 2 149.4 0 150.2 0
5 195.2 2 202.3 2 181.4 1 185.2 1 155.2 0 157.9 0
Table 4. Time and error data of (ABC) keyboard
Subjects
Session1 Session2 Session3 Session4 Session5 Session6
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
Time
(sec)
Error
letter
1 176.0 3 156.0 4 149.9 1 213.8 2 146.4 0 142.2 1
2 140.8 1 135.0 2 133.2 2 162.6 1 178.8 0 120.2 0
3 201.0 2 180.5 2 134.4 2 124.7 1 122.4 1 118.8 1
4 185.2 3 180.6 2 151.2 2 155.7 1 149.8 1 130.2 0
5 191.0 2 195.3 2 182.7 1 180.5 1 177.7 0 125.1 0
Figure 8. Text entry speed and error rate of (QWERTY) virtual keyboard

 ISSN: 1693-6930
926
Figure 9. Text entry speed and error rate of (ABC) virtual keyboard
4. CONCLUSION
The proposed virtual keyboard uses EEG signal synchronized with human-eye blinking in term of
key selection for printing purposes. The obtained results were encouraging and were about (1.5-1.8 WPM)
with an error rate equals to (5-5.25)%. The results of experiments show that the best mode is the one that
used the ABC keyboard type. The next challenge focuses on designing a similar keyboard with predicting
capabilities with text output interface for native languages.
ACKNOWLEDGEMENTS
The authors take this opportunity to express the profound gratitude and deep regard to
the presidency of the Northern Technical University for their constant scientific encouragement.
REFERENCES
[1] P. P. N. Ayudhya, S. Thaichinda, V. Saengkla, and P. Sittiprapaporn, “Electroencephalographic Study of Thai
Smoker’s Brain,” Int. ECTI North. Sect. Conf. Electr. Electron. Comp. And Telecommun. Eng, pp. 169–172, 2018.
[2] B. Kolb and R. Gibb, “Principles of Neuroplasticity and Behavior,” Cognitive Neurorehabilitation: Evidence and
Application, pp. 6–21, 2008.
[3] M. H. Masood, M. Ahmad, M. A. Kathia, R. Z. Zafar, and A. N. Zahid, “Brain Computer Interface Based Smart
Home Control Using Eeg Signal,” Sci.Int.(Lahore), vol. 28, no. 3, pp. 2219–2222, 2016.
[4] F. Mulla, E. Eya, E. Ibrahim, A. Alhaddad, R. Qahwaji, and R. Abd-Alhameed, “Neurological Assessment of
Music Therapy on the Brain using Emotiv Epoc,” 2017 Internet Technol. Appl. (ITA), pp. 259–263, 2017.
[5] M. Prabhakara and V. Kulkarni, “Real Time Analysis of EEG Signals on Android Application,” 2014 Int. Conf.
Adv. Electron. Comput. Commun. (ICAECC), 2014.
[6] W. Lawpradit and T. Yooyativong, “The EEG Brain Signal Representation for Surfaces and Shapes Touching
Behavior with an Inexpensive Device,” ECTI-NCON, pp. 135–140, 2018.
[7] N. K. Al-qazzaz, et al., “Role of EEG as Biomarker in the Early Detection and Classification of Dementia,”
The Sci. World J., vol. 2014, pp. 1-16, 2014.
[8] Y. J. Song and F. Sepulveda, “A Novel Technique for Selecting EMG-Contaminated EEG Channels in Self-Paced
Brain-Computer Interface Task Onset,” IEEE Trans. Neural Sys. Rehabil. Eng, vol. 26, no. 7, pp. 1353-1362, 2018.
[9] S. M. Hosni, H. A. Shedeed, M. S. Mabrouk, and M. F. Tolba, “EEG-EOG based Virtual Keyboard: Toward
Hybrid Brain Computer Interface,” Neuroinformatics, 2018.
[10] Mangala Gowri S. G., Dr. Cyril Prasanna Raj P., and Badarinarayan K. S., “Novel Algorithm for Feature Extraction
and Classification of EEG Signals,” Int. J. Eng. Res., vol. 4, no. 12, pp. 228–234, 2015.
[11] Y. Wen and Z. Huang, “Online Motor Imagery BCI Based on Adaptive and Incremental Linear Discriminant
Analysis Algorithm,” 9th
IEEE Int. Conf. Commun. Softw. Networks, (ICCSN), vol. 2017, pp. 962-966, 2017.
[12] M. Gaillard, S. Cognitives, E. Normale, and S. De Lyon, “Invasive and Non-Invasive Technologies in
Neuroscience Communication,” Biothique Online, pp. 1–10, 2017.
[13] S. Pb, S. Ts, T. Paul, E. P. John, and S. Peter, “Brain Computer Interface for Smart Home Control,” Int. J. of Advan
Research in Electrical, Electro and Instrumen Enginee., vol. 5, no. 4, pp. 170–177, 2016.
[14] F. Akram, M. K. Metwally, H. S. Han, H. J. Jeon, and T. S. Kim, “A Novel P300-based BCI System for Words
Typing,” 2013 Int. Winter Work. Brain-Computer Interface, pp. 24-25, 2013.

927
[15] J. Zhuang and G. Yin, “Motion Control of a four-wheel-independent-drive electric vehicle by motor imagery EEG
based BCI system,” Chinese Control Conf., pp. 5449-5454, 2017.
[16] A. Turnip, D. Soetraprawata, M. Turnip, and E. Joelianto, “EEG-based brain-controlled wheelchair with four
Different Stimuli Frequencies,” Internetworking Indones. J., vol. 8, no. 1, pp. 65–69, 2016.
[17] R. Raj, S. Deb, and P. Bhattacharya, “Brain Computer Interfaced Single Key Omni Directional Pointing and
Command System: A Screen Pointing Interface for Differently-abled Person,” Procedia Comput. Sci., vol. 133,
pp. 161-168, 2018.
[18] L. Zhang, Q. Lv, and Y. Xu, “Single Channel Brain-Computer Interface Control System Based on TGAM
Module,” Proc.-2017 10th
Int. Congr. Image Signal Process. Biomed. Eng. Informatics, vol. 2018, pp. 1–5, 2018.
[19] M. H. Hasbulah, F. A. Jafar, M. H. Nordin, and K. Yokota, “Brain-Controlled for Changing Modular Robot
Configuration by Employing Neurosky’s Headset,” Int. J. Adv. Comp. Sci. Appl, vol. 10, no. 6, pp. 114-120, 2019.
[20] R. A. Ramadan and A. V. Vasilakos, “Brain Computer Interface: Control Signals Review,” Neurocomputing,
vol. 223, pp. 26-44, 2017.
[21] J. Katona, T. Ujbanyi, G. Sziladi, and A. Kovari, “Speed Control of Festo Robotino Mobile Robot using NeuroSky
MindWave EEG Headset Based Brain-Computer Interface,” IEEE Int. Conf. Cogn. Infocommu, pp. 251-256, 2017.
[22] D. Martinez-maradiaga, and G. Meixner “Morpheus Alert: A Smartphone Application for Preventing
Microsleeping with a Brain-Computer-Interface,” 4th
International Conference on Systems and Informatics (ICSAI),
pp. 137–142, 2017.
[23] E. C. Djamal, D. P. Gustiawan, and D. Djajasasmita, “Significant Variables Extraction of Post-Stroke EEG Signal
using Wavelet and SOM Kohonen,” TELKOMNIKA Telecommunication Computing Electronics and Control,
vol. 17, no. 3, pp. 1149-1158, 2019.
[24] H. Fauzi, et al., “Energy Extraction Method for EEG Channel Selection,” TELKOMNIKA Telecommunication
Computing Electronics and Control, vol. 17, no. 5, pp. 2561-2571, 2019.
[25] V. I. Cherian, “Analysis of Text Entry Performance Metrics” Ahmed, J. Soc. Psychol., vol. 132, no. 2,
pp. 271–273, 2009.
BIOGRAPHIES OF AUTHORS
Dr.Thair Ali Salih received the MSc. Degree in Communication engineering from
the Technology University in 1986.He received the Ph.D. degree in Communication from
the Aleppo University in 2010. Currently, he is a Lecturer at Technical College/Mosul.
His research interests include spread spectrum systems and robotic system
Yasir Muslih Abdal, is a final year MSc. student, pursuing his Master of Technology
(M.Tech) Degree in Computer Technology Engineering at the Technical Engineering College
of Mosul Northern Technical University. His research interests revolved around bio-signal
processing and computer-brain interface.

Brain computer interface based smart keyboard using neurosky mindwave headset

More Related Content

What's hot (20)

Similar to Brain computer interface based smart keyboard using neurosky mindwave headset (20)

More from TELKOMNIKA JOURNAL (20)

Recently uploaded (20)

Brain computer interface based smart keyboard using neurosky mindwave headset