Emotiv EEG Classification System to

Determine the Motor Speed of a Wheelchair

Agus Siswoyo

and Ignatius Deradjad Pranowo

Mekatronika, Fakultas Vokasi Universitas Sanata Dharma, Yogyakarta, Indonesia

Keywords: BCI, Classification, EEG, Emotiv, Fuzzy Logic.

Abstract: This paper describes the use of the 14-channel Emotiv as a low-cost method of acquiring raw EEG signals,

for human-machine interfaces as a possible aid for persons with disabilities. To demonstrate the feasibility of

using raw EEG signals obtained using the Emotiv system, a classification algorithm based on Fuzzy logic is

implemented as an example. The proposed algorithm has been found to be effective in detecting and

classifying brain signals which can then be translated into valid commands for the human-machine interface.

The performance of the proposed approach is studied using fuzzy logic. The results obtained indicate a high

level of classification accuracy, therefore, indicating that the Emotiv EPOC can be used as a valid tool for

research.

1 INTRODUCTION

The research presented here is a continuation of

research that has been carried out using a type of

neurosky mindset sensor which is a Brain Computer

Interface (BCI) (Siswoyo, Arief, & Sulistijono,

2015). The BCI refer to a type of system that

combines electroencephalographic (EEG) activity

measurement technology with computational

development to convert brain activity into real

applications. The EEG sensor type Emotiv 14-

channel, which has a difference in the number of

points on the electrodes, is used in this study. This

study maps study data by application (BCI, signal

processing, experimental research, and validation)

and location of signal use.

The interest in BCI has mainly focused on medical

applications (Mahajan & Bansal, 2017); (Saifuddin

Saif, Ryhan Hossain, Ahmed, & Chowdhury, 2019);

(Cincotti et al., 2008). However, due to the increasing

desire of scientists and engineers to develop new

technologies, it is now also possible to find BCI in

applications such as video games, vehicle

manipulation, and psychological research. Related

work explaining how BCI can be used to control

drones to help people with disabilities can be found

https://orcid.org/0000-0002-4853-2191

https://orcid.org/0000-0002-0869-1761

in (Marin, Al-Battbootti, & Goga, 2020); (Abiyev,

Akkaya, Aytac, Günsel, & Çaǧman, 2016). A review

paper (Alotaiby, El-Samie, Alshebeili, & Ahmad,

2015), discussed some EEG channel selection

techniques for different applications taking into

consideration the different criteria developed in the

literature for channel selection evaluation. Of course,

there are many possibilities for BCI-EEG

applications (Banach, Małecki, Rosół, & Broniec,

2021); (Megalingam, Thulasi, & Krishna, 2013).

The classification of EEG signals is of significant

importance in BCI systems. Aiming to achieve

intelligent classification of EEG, a classification

methodology using sparse representation and fast

compression residual convolutional neural networks

is proposed (Huang, Li, Chen, Lin, & Yao, 2020).

This study propose an EEG classification system of

rule-based Emotiv Epoc signal output to identify

signals for moving wheelchair speeds. Therefore, we

believe that applying the Emotiv Epoc signal

classification to wheelchair speed movements, has

better results than the signals obtained from the

neurosky mindset.

Siswoyo, A. and Pranowo, I.

Emotiv EEG Classiﬁcation System to Determine the Motor Speed of a Wheelchair.

DOI: 10.5220/0011815300003575

In Proceedings of the 5th International Conference on Applied Science and Technology on Engineering Science (iCAST-ES 2022), pages 483-487

ISBN: 978-989-758-619-4; ISSN: 2975-8246

483

2 MATERIALS AND METHODS

2.1 EEG Data Recording

The raw EEG data were obtained from the 14-channel

electrode EEG data record of Emotiv Epoc (Figure 1).

The EEG is recorded when the test user thinks about

certain mental commands by imagining the

movement of objects. The test subject must be

focused, relaxed, and not have many distracting

movements when performing mental commands. The

experiment uses the "Mental Command Suite" to help

test subjects perform mental commands that allow

test subjects to control the movement of a 3-

dimensional cube using their minds. The 3D cube can

move up, down, left and right according to the mind

of the test subject.

Figure 1: Emotiv Epoc 14 channel.

Figure 2: Location of 14 Emotiv electrode channels.

2.2 Test Procedure and Environment

The use of the sensor is first given a saline solution

on the foam which is located at each electrode (Figure

2). Then do a check using the built-in software for the

connectivity of the electrodes with the location on the

scalp. When ready, a red colour will appear on the

indicator in the default software application.

2.3 EEG Data Analyses

All Emotiv 14 channel EEG data files were processed

and analyzed with MATLAB software (The

Mathworks, Inc.).

Figure 3: Experiment and data analysis.

EEG data preprocessing gets input from reading

the EEG signal which then forwards it to data

classification as shown in Figure 3 (Alabboudi,

Majed, Hassan, & Nassif, 2020). The processed raw

EEG data files are first labelled for each trial as

number "1" representing "slow wheelchair speed

trial" then number "2" representing "moderate

wheelchair speed trial" and number "3" represents "a

fast wheelchair speed trial"

The experimental results of this study continue

processing on the filtered data. EEG signal has weak

time-frequency-spatial characteristics, non-

stationary, non-linear, and weak intensity, so to

extract adaptive features reflecting frequency and

spatial characteristics, it is very important to adopt

feature extraction method. For this study, we

converted the time domain EEG data into the

frequency domain of the segments converted into

their respective frequency domains.

3 CLASSIFICATION MODELS

3.1 Classification Models and Metrics

When trying to collect signal data from Emotiv, data

will be displayed on the Emotiv default application.

There are 14 brain signal outputs, namely signals,

AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4,

F8, and AF4.

Figure 4: Run setup during data acquisition.

iCAST-ES 2022 - International Conference on Applied Science and Technology on Engineering Science

484

The subjects were asked to sit and comfortable

armchair with their upper limbs placed on the

armrests of the armchair (Figure 4). The EEG signal

Emotiv Epoc device was placed in each subject, some

of the technical descriptions include a sampling

frequency of 128 Hz, this device has 14 electrodes

and two ground references, and then data distributed

to the computer as shown in the Figure 5.

Figure 5: Emotiv Epoc output.

Figure 6: Signal F8, Frontal.

From the experimental results users, observing the

output of 14 signals, the F8 signal gives a significant

response when wheelchair users think to increase

speed (Figure 6).

Figure 7: Signal AF3, Frontal.

Then the AF3 signal sometimes appears to have a

significant spike in response at certain times of the

user (Figure 7).

The data totalling 14 channels is sent to

MATLAB, and then the data will be processed to

MATLAB (Figure 8).

Figure 8: Data output Emotiv in Matlab.

Classification of brain signals using fuzzy logic

(Siswoyo, Arief, & Sulistijono, 2017). From the

results of the Emotiv signal reading data, then the

signal data will be fuzzy. The research methodology

uses Fuzzy Logic Controller in this study Fuzzy used

the mamdani method, the input will be processed to

get the F8 value from the user, and this value is used

as a reference.

A complete fuzzy system consists of three main

components, namely:

1) Fuzzification

Fuzzification is a process of mapping input values

(crisp input) from a controlled system (non-fuzzy

quantities) into fuzzy sets according to their

membership functions from the emotive sensor itself.

To change the crisp input to fuzzy input, you must

first determine the membership function for each

crisp input, then the fuzzification process will take the

crisp input and compare it with the existing

membership function to generate fuzzy input values.

Figure 9: Membership fuzzy input Signal F8.

The F8 signal size indicates the level of focus. The

value ranges from 0 to 1000 (Figure 9). The level of

Emotiv EEG Classiﬁcation System to Determine the Motor Speed of a Wheelchair

485

mind focus increases when the user focuses his/her

mind.

Figure 10: Membership fuzzy input Signal AF3.

The size of the AF3 signal indicates the level of

focus. The values range from 0 to 1000 (Figure 10).

The level of focus increases when the user focuses

his/her mind. Figure 11 shows a variable output with

a range from 0 to 255.

Figure 11: Membership fuzzy wheelchair.

2) Design Rule Base

Here is the design rule of fuzzy logic that we can use

as shown in Table 1.

Table 1: Design rule of fuzzy logic.

Rules

F8 Signal

AF3 Signal

Low Middle High

Slow Minimum

Averag

Maximu

Normal Average

Averag

Average

Quick Minimum

Averag

Average

3) Defuzzification

From Table 1 we can get the defuzzification as:

a) If (F8 is LOW) and (AF3 is SLOW) then

(SpeedMotor is MINIMUM)

b) If (F8 is LOW) and (AF3 is NORMAL) then

(SpeedMotor is AVERAGE)

c) If (F8 is LOW) and (AF3 is QUICK) then

(SpeedMotor is MINIMUM)

d) If (F8 is MIDDLE) and (AF3 is SLOW) then

(SpeedMotor is AVERAGE)

e) If (F8 is MIDDLE) and (AF3 is NORMAL) then

(SpeedMotor is AVERAGE)

f) If (F8 is MIDDLE) and (AF3 is QUICK) then

(SpeedMotor is AVERAGE)

g) If (F8 is HIGH) and (AF3 is SLOW) then

(SpeedMotor is MAXIMUM)

h) If (F8 is HIGH) and (AF3 is NORMAL) then

(SpeedMotor is AVERAGE)

i) If (F8 is HIGH) and (AF3 is QUICK) then

(SpeedMotor is AVERAGE)

Nine rules were created for the system controller

to make up the rule base. The use of this fuzzy logic

method will follow a trial and error model.

4 CONCLUSIONS

Table 2 summarizes the selection criteria for the type

of EEG sensor for comparative analysis. Selection

criteria including the accuracy, sampling rate, ease of

use, number of channel, software application used,

communication method, learnability, and

performances.

Table 2: EEG Sensor type comparison.

No Compare

Neurosky

Mindset

Emotiv

Epoc

1 Accuracy Moderate High

2 Sampling rate 512Hz 128 Hz

3 Ease of use Easy Easy

4 Signal input channel 4 14

5 Software application Open Licence

6 Communication Bluetooth Wireless

7 Learnability Moderate Easy

8 Performance Moderate High

New users of Emotiv Epoc can gain control over

a single action fairly quickly. Learning to control

multiple actions usually takes practice and becomes

increasingly difficult as additional actions are added.

The user learns to train a reproducible mental state for

each action; detection becomes more and more

iCAST-ES 2022 - International Conference on Applied Science and Technology on Engineering Science

486

precise. Most users usually achieve their best results

after practicing each action several times. Practice

and experience will help determine the ideal amount

of training required for each individual user to

successfully control wheelchair speed.

In this research, a two input from signal Emotiv

Epoc sensor, one output MFIS in fuzzy tool box of

Matlab software was used for control speed

wheelchair. Grading results obtained from fuzzy logic

showed a good general agreement (91%) with the

results from the human experts, providing good

flexibility in reflecting the expert expectations and

grading standards into the results. This model

demonstrated that, control speed evaluation based on

this method is more exact than experts, and provides

a better representation control speed grading.

Another topic for future work is the effectiveness

of EEG signals used for the needs of people with

disabilities. Different users allow different responses

to the same stimulus. Ease of extracting task-relevant

EEG patterns from recordings signal.

ACKNOWLEDGEMENTS

This research was fully funded by the Academic

Directorate of Vocational Higher Education,

Directorate General of Vocational Education

Ministry of Education Culture Research and

Technology, Fiscal Year 2022 (SP DIPA-

023.18.1.690524/2022) with contract No.

127/SPK/D.D3/PPK.01.ATVP/VI/2022, and also

supported by LPPM Sanata Dharma University (No.

031 Penel./LPPM-USD/VII/2022).

REFERENCES

Abiyev, R. H., Akkaya, N., Aytac, E., Günsel, I., &

Çaǧman, A. (2016). Brain-Computer Interface for

Control of Wheelchair Using Fuzzy Neural Networks.

BioMed Research International, 2016. https://doi.org/

10.1155/2016/9359868

Alabboudi, M., Majed, M., Hassan, F., & Nassif, A. B.

(2020). EEG wheelchair for people of determination. In

2020 Advances in Science and Engineering Technology

International Conferences, ASET 2020. https://doi.

org/10.1109/ASET48392.2020.9118340

Alotaiby, T., El-Samie, F. E. A., Alshebeili, S. A., &

Ahmad, I. (2015). A review of channel selection

algorithms for EEG signal processing. Eurasip Journal

on Advances in Signal Processing, 2015(1).

https://doi.org/10.1186/s13634-015-0251-9

Banach, K., Małecki, M., Rosół, M., & Broniec, A. (2021).

Brain-computer interface for electric wheelchair based

on alpha waves of EEG signal. Bio-Algorithms and

Med-Systems, 17(3), 165–172. https://doi.org/

10.1515/bams-2021-0095

Cincotti, F., Mattia, D., Aloise, F., Bufalari, S., Schalk, G.,

Oriolo, G., … Babiloni, F. (2008). Non-invasive brain-

computer interface system: Towards its application as

assistive technology. Brain Research Bulletin, 75(6),

796–803. https://doi.org/10.1016/j.brainresbull.

2008.01.007

Huang, J. S., Li, Y., Chen, B. Q., Lin, C., & Yao, B. (2020).

An Intelligent EEG Classification Methodology Based

on Sparse Representation Enhanced Deep Learning

Networks. Frontiers in Neuroscience, 14(September),

1–8. https://doi.org/10.3389/fnins.2020.00808

Mahajan, R., & Bansal, D. (2017). Real Time EEG Based

Cognitive Brain Computer Interface for Control

Applications via Arduino Interfacing. Procedia

Computer Science, 115, 812–820. https://doi.org/

10.1016/j.procs.2017.09.158

Marin, I., Al-Battbootti, M. J. H., & Goga, N. (2020). Drone

Control based on Mental Commands and Facial

Expressions. In Proceedings of the 12th International

Conference on Electronics, Computers and Artificial

Intelligence, ECAI 2020 (pp. 1–4).

https://doi.org/10.1109/ECAI50035.2020.9223246

Megalingam, R., Thulasi, A., & Krishna, R. (2013).

Thought Controlled Wheelchair Using EEG

Acquisition Device. In 3rd International Conference on

Advancements in Electronics and Power Engineering

(ICAEPE’2013) (pp. 207–212). Retrieved from

http://psrcentre.org/images/extraimages/113543.pdf

Saifuddin Saif, A. F. M., Ryhan Hossain, M., Ahmed, R.,

& Chowdhury, T. (2019). A Review based on Brain

Computer Interaction using EEG Headset for

Physically Handicapped People. International Journal

of Education and Management Engineering, 9(2), 34–

43. https://doi.org/10.5815/ijeme.2019.02.04

Siswoyo, A., Arief, Z., & Sulistijono, I. A. (2015). A

Design of Brain Computer Interface Based Fuzzy Logic

for Control of Motor Speed. In The Fourth Indonesian-

Japanese Conference on Knowledge Creation dan

Intelligent Computing (KCIC) (p. ISBN 978-602-

72251-0-7, 2015.). Surabaya, Indonesia: Politeknik

Elektronika Negeri Surabaya.

Siswoyo, A., Arief, Z., & Sulistijono, I. A. (2017).

Application of Artificial Neural Networks in Modeling

Direction Wheelchairs Using Neurosky Mindset

Mobile (EEG) Device. EMITTER International Journal

of Engineering Technology, 5(1), 170–191.

https://doi.org/10.24003/emitter.v5i1.165.

Emotiv EEG Classiﬁcation System to Determine the Motor Speed of a Wheelchair

487