Machine learning in human computer interface

Machine Learning in Human Computer Interface
R. Rajkumar
Department of Computer Science and Engineering
School of Computing
SRM Institute of Science and Technology
AICTE SPONSORED SIX DAYS ONLINE SHORT TERM TRAINING PROGRAM (STTP) ON
“DATAANALYTICS AND SUPPORTING SERVICES USING MACHINE LEARNING”

01-12-2020 R. Rajkumar SRM IST | Chennai 2
Human Computer Interface

Brain Computer Interface
• Brain-computer interface (BCI) is a fast-growing emergent technology in
which researchers aim to build a direct channel between the human brain
and the computer.
BCI- The
Ultimate in
Human
Computer
Interfacing

6
Brain-Computer Interface (BCI)
• A Brain-Computer Interface (BCI) is a device that enables
communication without movement
• A Brain Computer Interface, is a direct communication pathway
between human or animal brain and an external device.
• It usually connects the brain with a computer system.
• BCI can help people with inabilities to control computers.
01-12-2020 R. Rajkumar SRM IST | Chennai

WHY BCI?
BCI can control
wheelchairs,
televisions,
or other devices.

THOUGHTS faster than ACTIONS
Actions are mechanical
movements while thoughts are
electrical impulses.
Thoughts are faster than actions
So direct interaction with the world
through thoughts would be faster.

Usage of BCI
• BCI is a new neuroscience paradigm that might help us
better understand how the human brain works
• BCI research allows us to develop a new class of
bioengineering control devices and robots
• BCI hold promise for rehabilitation and improving
performance
• BCI can expand possibilities for advanced human computer
interfaces (HCIs)

10
BCI Communication
Based on the communicative Pathway BCI is classified as follows
 One Way BCI
- Computers accept commands from the brain
 Two Way BCI
- Brains and external devices can exchange information in both
directions.

11
Achievements

R. Rajkumar SRM IST | Chennai 12
Brain Gate Research in Human
• Mathew-Nagel the first person to use the brain-
computer interface to restore functionality lost due
to paralysis…
01-12-2020

Honda Asimo
• Honda has demonstrated a person sending four
simple commands to the robot simply by thinking.
01-12-2020

Remote controlled Rat or RATBOTS
• Guided rats controlled through implants in their
brains could one day be used to search for landmines
or buried victims of earthquakes,
01-12-2020

Bionic eye
01-12-2020

What is it good for?
• Neurofeedback
• Treating attention deficit hyperactivity disorder (ADHD), poor concentration
• Brain Computer Interfaces
• People with little muscle control (i.e. not enough control for EMG or gaze
tracking)
• People with spinal injuries
• High Precision
• Low bandwidth (bit rate)

Area of implementation
General application
Control a robot
Playing games
For physically weak persons to handle the computer
Cursor control
o Allow those with poor muscle control to
o communicate and control physical devices

Area of implementation
Medical science
 Enabling disabled people
o Vision and hearing
o Paralysis treatment
o Prosthetic devices (legs, hands etc)
o Provide a means of communication to completely paralyzed patients
o Surgically implanted devices used as replacement
for paralyzed patients.
R. Rajkumar SRM IST | Chennai 1801-12-2020

HUMAN AND THE WORLD
Human interacts with the world using his five senses

20
We are going to study about
the most complex living
structure on the universe

• About 3 pounds (1,300 grams)
• 78% water, 10% fat, 8% protein
• Less than 2.5% of body’s weight
• Uses 20% of body’s energy
• People only use 10% of their brain
• 100 billion neurons (Greek word meaning
bowstring)
• 1 trillion glial cells (Greek word meaning glue)
• 1,000 trillion synaptic connection points
• 280 quintillion memories
Facts about Brain

• Brain is modified by environment
• Brain is adaptable Plasticity ‘Use it or Lose it’
• Left and Right Hemispheres
• Left hemisphere for most people is the dominant hemisphere-
responsible for production of language, mathematical ability,
problem solving, logic
• Right hemisphere thought to be responsible for creativity and
spatial ability
• Everything people have ever experienced is stored somewhere
in their brain
Facts about Brain

Two Kinds of Brain Cells
•Glia - (Greek word meaning glue)
•90% of the brain cells
•Less known about glia cells
•No cell body
•Remove dead brain cells and give
structural support

Two Kinds of Brain Cells
• Neurons (Greek word meaning
bowstring)
• 100 billion neurons in human brain
• Neurons essential to performing the
brain's work
• Consist of a compact cell body,
dendrites, and axons

Neurons
• Neurons (brain cells) make connections between
different parts of the brain.
• Information is carried inside a neuron by electrical
pulses and transmitted across the synaptic gap
from one neuron to another by chemicals called
neurotransmitters.
• Learning is a critical function of neurons.

Dendrites and Axons
• Dendritic branching helps make connections
between cells.
• As cells connect with other cells, synapses occurs.
• New synapses appear after learning.
• Repeating earlier learning makes neural pathways
more efficient through myelination
(fatty substances formed around axons, Myelination
enables nerve cells to transmit information faster and
allows for more complex brain processes)

The nerve cell, or neuron Synaptic Density
2
year
old
6
year
old

Lobes of the Brain
• Frontal
• Parietal
• Occipital
• Temporal

Lobes of the Brain - Frontal
• The Frontal Lobe of the brain is located deep to the
Frontal Bone of the skull.
• It plays an integral role in the following
functions/actions:
- Memory Formation
- Emotions
- Decision Making/Reasoning
- Personality

Lobes of the Brain - Parietal Lobe
• The Parietal Lobe of the brain is located deep to
the Parietal Bone of the skull.
• It plays a major role in the following functions/actions:
- Senses and integrates
sensation(s)
- Spatial awareness and
perception

Lobes of the Brain – Occipital Lobe
• The Occipital Lobe of the
Brain is located deep to the
Occipital Bone of the Skull.
• Its primary function is the
• Processing,
• Integration,
• Interpretation, etc. of VISION
and visual stimuli.

Lobes of the Brain – Temporal Lobe
• The Temporal Lobes are located on the sides of
the brain, deep to the Temporal Bones of the skull.
• They play an integral role
in the following functions:
- Hearing
- Organization/Comprehension
of language
- Information Retrieval
(Memory and Memory
Formation)

• Two sides or hemispheres of the brain: LEFT and RIGHT
• We have two cerebral hemispheres connected by the
corpus callosum. This is a bundle of nerves that allows
each side of the brain to communicate with each other.
• Each side of the brain processes things differently.
Right & Left Brain

How the two sides process information that is!
Left Brain
• Logical
• Sequential
• Rational
• Analytical
• Objective
• Looks at parts
Right Brain
• Random
• Intuitive
• Holistic
• Synthesizing
• Subjective
• Looks at wholes

Left Hemisphere
• Processes things more in parts and sequentially
• Recognizes positive emotions
• Identified with practicality and rationality
• Understands symbols and representations
• Processes rapid auditory information faster than the right
(crucial for separating the sounds of speech into distinct
units for comprehension)
• is responsible for language development. It develops
slower in boys, that is why males usually develop more
language problems than females.

Right Hemisphere
• Recognizes negative emotions
• High level mathematicians, problem solvers, and chess
players use
• Responds to touch and music (sensory)
• Intuitive
• Responsive to color and shape
• Emotional and originative

What information the two sides recognize!
Left Brain
• Letters
• Numbers
• Words
Right Brain
• Faces
• Places
• Objects

Activities
Auditory ActivityVisual ActivityThinking Activity
Memory Activity Motor Activity Seeing, ………

41
HOW BCI WORKS
•Electrodes collect electric signals or electrical
impulses and patterns.
• Amplify the signal approximately ten thousand
times
•Turn these signals to the machine
understandable language (into some 0’s and 1’s)
•Then computer then translates these electric
signals into Human understandable data.

• Invasive techniques, which implant electrodes
directly onto a patient’s brain
• Noninvasive techniques, in which medical
scanning devices or sensors mounted on
caps or headbands read brain signals (EEG)
• Partial Invasive Techniques
How to read Brain Signals

43
Block Diagram of BCI

44
Invasive BCIs
• Invasive BCIs are implanted directly into
the brain by neurosurgery.
• Produce the highest quality signals of BCI
devices.
• But are prone to scar tissue build-up.
Glial cells provide support and protection
for neurons and electric signals.

45
Invasive BCIs

46
Partially Invasive
• It is applied to the inside the skull but outside the grey matter.
• It lowers risk of forming scar-tissue in the brain

Non-Invasive BCIs

48
• It is the most useful neuron signal acquisition
method which is applied to the outside of the
skull, just applied on the scalp.
Non-Invasive BCIs

PHYSICAL MECHANISMAS
• EEGs require electrodes
attached to the scalp with
sticky gel
• Require physical connection
to the machine

Electrode Placement
• Standard “10-20 System”
• Spaced apart 10-20%
• Letter for region
• F - Frontal Lobe
• T - Temporal Lobe
• C - Center
• O - Occipital Lobe
• Number for exact position
• Odd numbers - left
• Even numbers - right

Electrode Placement
• A more detailed view:

EEG - Electro encephalography
ECoG - Electro cortico graphy
MEG - Magneto encephalo graphy
BOLD - Blood-Oxygen-Level-Dependent (signal)
MRI - Magnetic Resonance Imaging
Different neuroimaging methods are used to derive
meaningful interpretations from the brain signals which
are captured by microelectrodes:
Signal extraction techniques

WHAT IS AN EEG?
• An electroencephalogram is a measure of the brain's
voltage fluctuations as detected from scalp electrodes.
• It is an approximation of the cumulative electrical activity
of neurons.
• EEG measures the electrical activity of the brain with
multiple electrodes placed on the scalp
• It is portable and relative inexpensive that is why most
noninvasive BCIs are presently based on EEG.

EEG Background
• 1875 - Richard Caton discovered electrical properties of exposed cerebral
hemispheres of rabbits and monkeys.
• 1924 - German Psychiatrist Hans Berger discovered alpha waves in humans
and invented the term “electroencephalogram”
• 1950s - Walter Grey Walter developed “EEG topography” - mapping
electrical activity of the brain.

EEG Technique for BCI
Electrodes attached to the scalp
Electric signals of the brain are amplified.
Transmitted to the computer
Software converts them into technical control
signals ( computer commands)
These computer commands controls the devices.

EEG WAVES
There are 5 major types of EEG waves
1. Delta Waves
2. Theta Waves
3. Alpha Waves
4. Beta Waves
5. Gamma Waves

Continuous Brain Waves
Generally grouped by frequency: (amplitudes are about 100µV max)
Type Frequency Location Use
Delta <4 Hz everywhere occur during sleep, coma
Theta 4-7 Hz temporal and parietal correlated with emotional stress
(frustration & disappointment)
Alpha 8-12 Hz occipital and parietal reduce amplitude with sensory
stimulation or mental imagery
Beta 12-36 Hz parietal and frontal can increase amplitude during intense
mental activity
Mu 9-11 Hz frontal (motor cortex) diminishes with movement or intention
of movement
Gamma 36< Hz temporal Multi tasking activities

Alpha and Beta Waves
• Studied since 1920s
• Found in Parietal and Frontal Cortex
• Relaxed - Alpha has high amplitude
• Excited - Beta has high amplitude
• So, Relaxed -> Excited
means Alpha -> Beta

Mu Waves
•Studied since 1930s
•Found in Motor Cortex
•Amplitude suppressed by Physical
Movements, or intent to move physically
•(Wolpaw, et al 1991) trained subjects to
control the mu rhythm by visualizing motor
tasks to move a cursor up and down (1D)

Mu and Beta Waves
• (Wolpaw and McFarland 2004) used a linear
combination of Mu and Beta waves to control a 2D
cursor.
• Weights were learned from the users in real time.
• Cursor moved every 50ms (20 Hz)
• 92% “hit rate” in average 1.9 sec

Hardware Components
• A Neuro Chip
• Connector
• Converter &
• Computer
01-12-2020

62
The components in BCI system
• A 4-millimeter square silicon chip
• 100 hair-thin microelectrodes
• Embedded in the primary motor cortex
- the region of the brain responsible
for controlling movement.
• Microelectrods detects electrical signals
generated when a user imagines.
The Neuro chip:

63
The connector:
The signal from the brain is
transmitted through the pedestal
plug attached to the skull.

64
The converter:
•The signal travels to an amplifier
•These amplifiers amplify the signal
approximately ten thousand times
•It is converted to optical data and
bounced by fibre-optic cable to a
computer.

65
The computer:
The computer translates brain activity
and creates the communication using
custom decoding software.

BCI Examples - Prostheses
• (Wolpaw and McFarland 2004) allowed a user to move a
cursor around a 2 dimensional screen
• (Millán, et al. 2004) allowed a user to move a robot around
the room.

Student walking in
the virtual world with
the character
controlled by his
brain waves.
More examples

BCI Examples - Music
• 1987 - Lusted and Knapp demonstrated an EEG
controlling a music synthesizer in real time.

Example of BCI application
A physically
handicapped
man operates
a BCI
wheelchair

Deep Learning in Brain-Computer Interface

Break

Deep Learning in Brain-Computer Interface
Other than controlling devices, different applications and studies of BCIs are:
• Sleep pattern
• Attention deficit hyperactivity disorder (ADHD)
• Disorders of consciousness
• Depth of anaesthesia
• Fatigue and mental workload
• Mood
• Emotions

The current performance of deep learning models is unclear
whether it can outperform traditional processing techniques.
This is because unlike natural image where there is the ImageNet
dataset as a benchmark dataset, EEG does not have a benchmark
dataset.
Unfortunately for BCI studies, many researchers use private
dataset, Acquiring data is more expensive and annotated data
requires subject matter experts contributions.
Current state-of-the-art

As the challenges of BCI applications, EEG signals are highly variable. EEG
signals can differ significantly between subjects and even within the same
subject, as the EEG contains interferences from ongoing brain activity and
measurement noise.
These types of noise suggest the use of regularisation in order to keep the
weights of the network small to reduce overfitting.
The common regularisation methods for neural networks are L1 and L2,
which add a penalty to the weights according to their magnitude and sign.
Dropout technique is very commonly used as well.
These regularisation techniques generally improve the performance slightly
and most research used at least one regularisation techniques.
Regularisation

How Deep Learning Works with BCI
Deep learning has shown significant performance in various tasks; it
has proven to outperform “traditional” machine learning approaches
that use handcrafted features.
Decoding the brain electrical activity with high variability and non-
stationary noise into a meaningful signal is difficult.
These difficulties lead to the use of machine learning algorithms to
solve BCI applications.

How Deep Learning Works with BCI
Example

is it the Capital of Tamilnadu?
Coimbatore

Chennai

Madurai

Vellore

AaaBBaaaCC
BBccAAccA BbccAAccA CcccAAccA

Deep learning has the ability to extract features and learn from
hierarchical representations from high dimension data and has lead to
many real-world applications in the area of computer vision and natural
language processing.
Given its effectiveness in other fields, deep learning seems promising to
learn from raw EEG data to extract better features to improve
performance and robustness.
In order to tackle the challenges in BCI application, researchers work on
to improve extraction of essential features from EEG signals and to
build models that can generalise better.
How Deep Learning Works with BCI (Con)

Problems with deep learning in BCI
applications
• As training a deep learning model generally requires a large training
dataset. Unlike the computer vision research community, where a
vast amount of data is available; limited BCI data available pose a
challenge to advance the field forward. High-quality data is also
challenging to acquire, augmenting datasets or use of generative
adversarial networks might be adopted.
• Deep learning models are great in memorising dataset, but given that
EEG has low signals to noise ratio, models might memorise noise
data. As a result, the performance is greatly affected even with
various regularisation techniques.

Method: Restricted Boltzmann machine
A restricted Boltzmann machine (RBM) learns the probability
distribution of the input data based on a gradient ascent of the log-
likelihood of the training data.
Deep Belief Network (DBN) composed of three RBMs, where RBM can
be stacked and trained in a deep learning manner.

Method: Recurrent neural network
Given that EEG data has a temporal
structure, frequencies over time, the
recurrent neural network (RNN) is
suitable. RNN model sequential data via
recursive, which is unfolding the RNN in
time to form a feed-forward neural
network to apply backpropagation.
Long short term memory (LSTM) is an RNN architecture that is
composed of memory blocks which use gating units with a self-
connected memory cell. LSTM solves the vanishing gradient problem
that traditional RNNs suffer.

Contribution for BCI Research
Researchers need to develop more robust and consistent algorithms
that can be easily trained and deployed. Algorithms must be able to
work with small training samples, handle noisy signals, generalise well
on users over different time as well as mood and generalise well over
different users.
Many studies are evaluated offline on a small number of subjects, but
for actual BCI applications to work machine learning need to work in
real-time.

BCI Challenging Fields:
• Neuroscience
• Signal processing
• Machine learning
• Computational intelligence
• Cognitive science
• Physics

Case Study
Bio-Inspired Brain Computing Interface
Learning Style Inventory to Increase the
E-Learning Efficiency using Machine Learning

Growth of MOOC
01-12-2020 89R. Rajkumar SRM IST | Chennai

Synchronous E- Learning
This mode of e-learning delivery allows
real time interaction between both the
parties (learners and instructor). This
form of delivery could be in the form of
multicast web seminars (Webinars), chat
or even tele-video conferencing.

Asynchronous E- Learning
This mode of delivery does not allow real time
interaction between the student and instructor. This
form of delivery channels could be through Learning
Management Systems(LMS), and Massive Online Open
Courseware (MOOC).
Instructors and Course Developers must simultaneously
consider the design of e-learning courses to ensure
optimum impact.

E- Learning: Multidisciplinary approach
01-12-2020 92
4. Research
i. New Innovation
ii. Re-designing
iii. Usability and Acceptance
iv. Ongoing evaluation
v. Formal and informal assessment
vi. Design Issues
1. Pedagogical
2. Technology
3. Policy
4. Research
R. Rajkumar SRM IST | Chennai

How we learn: Difference between learners brain
Each Learner’s brain is
organized in a unique way.
Each learner has a complex
and dynamic profile of
strengths and limitations.

Learning Styles
01-12-2020 94
Researchers recognize that each person prefers
different learning styles and techniques.
Learning styles group common ways that people
learn.
Interestingly, there have been more efforts at
advancing technology than on attempting to
understand the needs and learning styles of
individual learners and instructional design.

01-12-2020 95
Learning Style Models with no of articles
VARK/VAK
Felder–Silverman
Kolb
Mixed
CognitiveStyle
Myers-Briggstype
Honey&Mumford
Gregore
Keefe
Wikkin&
goodenough
Undetemined

01-12-2020 96
Introvert and Extrovert Personality Traits

01-12-2020 97
INTROVERTS
EXTROVERTS
STIMULU
S
STIMULU
S
PROCESSED BY
BRAIN
PROCESSED BY
BRAIN
Introvert and Extrovert differences

Introvert and Extrovert Scientific Evidence
01-12-2020 98
Extroversion is a phenomenon, in which the human brain needs constant
stimulation and radiates energy in the form of intense emotions and feelings.
For introverts, stimulation travels through a long and a complicated pathway
through many areas of brain, including the right front Broca’s area – self-talk,
insular – empathy and Left hippocampus–personal.
Introverts may process information more thoroughly and deeply than
extroverts doing the thought processing.

Introvert and Extrovert Brain differences

Learners in E-Learning
01-12-2020 100
Auditory LearnersVisual Learners

Learners in E-Learning
01-12-2020 101
Auditory

Chatbot in E-Learning
01-12-2020 102
In the field of E-learning, the application of a Chatbot as part of the
education has shown interesting potential, both as a teaching and
administrative tool.
Chatbots have been ‘trending’ for a few years and quite a few papers
examining them in the educational sector have been published, albeit very
little interest seems to have been given to the summation of this
knowledge.

Chatbot in E-Learning

Problem Statements
01-12-2020 104
• Researchers are attempting to decrease the dropout rate of E-Learning courses
using various methods.
• A recent study, in online courses, it has been found [43] by Massachusetts
Institute of Technology (MIT) that an astronomical dropout rate of about 96 per
cent was found for the last five years.
• There is a need in Human Computer Interface (HCI) to use Brain Computer
Interface (BCI) with Learning Styles to increase the efficiency of the E-Learning.
• There is need to explore the possible method to group the learners into
personality traits and learning styles.

Motivation
of the
Research

Objectives
1.To use Chatbot instead of survey questionnaires for E-
learning.
2.To record brain signals using Brain Computing Interface
for learner’s Classification.
3.To find out the correlation between Learning styles and
Personality traits using machine learning algorithms.

01-12-2020 107
Existing Method: VARK / VAK Learning Style

Proposed Chatbot (Bio - Inspired)
The proposed Chatbot has loop of
questions that are divided into three parts,
1. Learners’ general likes and dislikes
2. Learners’ day to day unique habits
3. Learning preferences on learning

Method
Initially, modified VARK questionnaires are implemented as a Chatbot to
classify the individuals as Visual or Auditory Learners.
The Chatbot is intended to know the learners’ preferences manually. The
conversation of the Bot in Chatbot is likely to interact with the learners to
know the basic information like age group, habits and learning preferences.
The Chatbot interacts with the learners to classify them into Introverts and
Extroverts.

Proposed Bio Inspired
Chatbot Flow chart
‘

Experimentation
The experimentation of Chatbot is conducted for different age group of
learners.
The Chatbot’s Unified Resource Locator (URL)
https://sites.google.com/view/bcibli/home
is shared with 118 learners from SRM Institute of Science and Technology,
Kattankulathur, Chennai, India.
As per the data flow of the proposed Chatbot, the implementation is
performed using Landbot online tool which is a tool for creating online
interfaces for Chatbot.

Chatbot Classification: Results
01-12-2020 113
SI.NO
CLASSIFICATI
ON
COUNTS
Average
time taken
1 Extroverts 56
3.53
minutes
2 Introverts 47
3
Not to
classify
(Ambiverts)
15
Total 118

To record brain signals using
Brain Computing Interface for
learner’s Classification

01-12-2020 115
Experimentation

BRAIN COMPUTER INTERFACE FOR CLASSIFICATION OF
LEARNERS USING MACHINE LEARNING
Brain Computer Interface (BCI) is an advanced version of emerging technology
called Human Computer Interface (HCI).
Recently, non-invasive BCI sensor devices are commercially available for learning
context. BCI products like Muse [49] and Neuorosky [50] are used in meditation and
concentration training.
These devices are absolutely wire free. They can connect with Bluetooth
communication interface technology.

Hardware Used
The hardware of the BCI devices is able to connect with the mobile devices
like iPhone and Android phones.
Neurosky Mind wave Mobile 2 [51] is a single electrode BCI device which is
made use of in capturing the EEG signals for experiments in this study.

Details of the participants
A total of 118 healthy learners are chosen
for this experiment in the age groups
between 18 and 79 years. They are
requested to watch the sample audio and
video learning contents.
The learners participated in this study are
from SRM Institute of Science and
Technology located at Kattankulathur, India.

Preparation
The learners are seated in a silent atmosphere. As per the
international guidelines learners are requested to take a long
breath for at least ten times before they are allowed to record
their brain EEG Signals.

Tools for BCI
The receiver side of the mobile device is a mobile
application called EEG ID.
It is light weighted application and it can connect with
Neurosky device through Bluetooth connection.
EEG ID using the EEG signals can be recorded at
intervals of minutes, seconds and milliseconds.
The Learner can generate BCI waves and that will be
converted into Comma Space Value (CSV) file through
EEG ID Mobile application.

Sample Video Content
01-12-2020 121
The sample video contains Visual and Auditory
contents.
Learners are requested to watch the sample audio
and video learning contents for two minutes.

Brain Waves
01-12-2020 122
While watching the sample Audio Visual content,
each learner produces beta waves for every
second continuously.
Once the learners calm down themselves the
testing sample contents are shown one by one to
all the 118 learners. The Beta waves are
produced in the range from 180 Hz to 16776 KHz.

Observations:
01-12-2020 123
• It is observed that every learner had produced a unique EEG brain wave.
• During the testing, It is clear that Introverts produce a high magnitude of Beta
waves during the initial period when the visual contents are shown, whereas the
Extroverts produce low amplitude Beta waves during the initial period when the
visual contents were shown.
• Ambiverts produce lesser modulation of beta waves during the entire testing
period.

Normalization of Beta Waves
The EEG brain waves have quite large variations. EEG brain wave
datasets are to be normalized before proceeding for classification.
01-12-2020 124
Pseudo Code:
Read x
If (x <= 999)
Set x to 1;
Else if (x <= 99999)
Set x to 2;
Else if (x<= 999999)
Set x to 3;
Else
Set x to 4;
End

Data set after normalization
01-12-2020 125
Value Count
Total Instances 14, 396
Minimum value 1
Maximum value 4
Mean 2.407
Standard Deviation 1.262

To find out the correlation between Learning
styles and Personality traits using machine
learning algorithms.

Machine Learning for classification

Machine Learnig Tools Comparison

Machine Learning with WEKA
01-12-2020 129
In the experiments conducted, the BCI data set is time dependent.
Training : 15 – Introverts , 15 – Extraverts , 10 – Ambiverts
Testing : 118 Instances
Since the data sets generated are time dependent data, and Naïve Bayes and
J48 classification algorithms are applied to the data sets and a confusion
matrices are formed.

Classification using Naïve Bayes
01-12-2020 130
The BCI dataset of 118 learners is applied to the Naïve
Bayes classification algorithm available in WEKA tool.
Parameters Values
Correctly Classified Instances 110 (93.2203 % Accuracy)
Incorrectly Classified Instances 8 (6.7797%)
Mean absolute error 0.8872
Root mean squared error 0.0452
Relative squared error 0.2126
Relative absolute error 11.36%
Total Number of Instances 118

Classification using Naïve Bayes
01-12-2020 131
a b c Classified
as
50 3 3 a = E
0 48 0 b = I
0 2 12 c = A

01-12-2020 132
Classification using J48 Classifier Algorithm
Parameters Values
Correctly Classified Instances 112 (94.2203 % Accuracy)
Incorrectly Classified Instances 6 (5.0847 %)
Mean absolute error 0.0595
Root mean squared error 0.1724
Relative absolute error 14.94%
Root relative squared error 38.71%
Total Number of Instances 118

01-12-2020 133
Classification using J48 Classifier Algorithm
a b c Classified as
52 1 3 a = E
0 48 0 b = I
0 2 12 c = A

Comparison between proposed method and
machine learning classification

Procedures to avoid over-fitting
01-12-2020 135
Detecting over-fitting is useful. Cross-validation is a powerful
preventative measure against over-fitting.
The classic way to avoid over-fitting is to divide the data sets into
three groups a training set, a test set, and a validation set. To validate
the over-fitting clustering algorithm is applied by removing the labels
in dataset.

After applying the Canopy Clustering algorithm
01-12-2020 136
Clusters Total numbers
0 53 (45 %)
1 50 (42 %)
2 15 (13 %)
Number of Canopies found 3
T2 radius 4.597
T1 radius 5.746
Time to build the model
0.13
Seconds

Comparison
01-12-2020 137
Algorithm
Parameter
Proposed Bio-
Inspired
Chatbot Method
Naïve Bays
Classification
Algorithm
J48 Tree
Classification
Algorithm
Canopy
Clustering
Algorithm
Accuracy NA 93.2203% 94.9153% NA
Classification /
Clusters
E – 56 *
I – 47**
A-15 ***
E - 53*
I – 48 **
A. 14 ***
E – 52 *
I – 48**
A. 12 ***
E- 53*
I – 50**
B-15***
Not Classified Nil 8 6 Nil
*Extravert **Introvert ***Ambivert NA- Not Applicable

Conclusion
Machine Learning algorithms are used for validating the
classification of Learning Style Inventory with Personality traits of
Introverts and Extroverts.
The proposed Bio Inspired Chatbot method gives good accuracy in
classification of learners than the existing (VARK) method.
It has been observed that, Introvert learner prefers Visual
Contents and Extrovert Learner prefers Auditory contents from the
Brain Computer Interface Experiments.

139
Advantages of One way Communication
• Brain Controlled Devices
• Gamers to control games
• Mute person to speak
• Wheelchair Operation
• Robotic Arms
• Elder’s Care System

140
Disadvantages of Two way Communication
• Cost
• Research is still in beginning stages.
• Easy to hack the brain (information)
- Controllng one's thoughts
- Reading one's thoughts
• Risk for surgery failure
• Create scar tissue in the brain
• Security issues

• Enabling disabled people
o Vision and hearing
o Paralysis treatment
o Prosthetic devices (legs, hands etc)
• Psychotherapy
o Diagnostics
o Treatment
• Military and civil research
o Making dangerous jobs
BCI APPLICATIONS…

BCI APPLICATIONS CONT…
o Provide a means of communication to completely paralyzed
patients
o Surgically implanted devices used as replacement for
paralyzed patients
o Allow patients to control a computer by conscious changes
of brain activity
o Allow those with poor muscle control to
communicate and control physical devices

Levels of BCI

Conclusion
• BCI is highly promising.
• Provide high standard of living.
• Extend our limits.
• Impart a new level to the popular quote-
• “I think therefore I am!”

Future developments
• Better signal detection
• Shortening training time
• Improving learning (neurobiological and psychological basis)
• New recording methods (NIRS, ECoG)
• Broader range of applications (interface to commercially
available assistive devices; treatment of diseases)

BCI Lab

Journal Publications
1. R. Rajkumar, V. Ganapathy, “BIO-INSPIRING LEARNING STYLE CHATBOT INVENTORY USING BRAIN
COMPUTING INTERFACE TO INCREASE THE EFFICIENCY OF E-LEARNING”, IEEE – ACCESS. MARCH, 2020, ID
ACCESS-2020-13266.
2. R. Rajkumar, V. Ganapathy, “BIO INSPIRED BLOOD GROUP PREDICTION”, International Journal of Control
Theory and Applications. Volume 13. No 04. 112- 117. June 2017.
3. R. Rajkumar, V. Ganapathy, “BRAIN COMPUTER INTERFACE TO IDENTIFY THE STATE OF MIND OF THE
LEARNERS”, International Journal of Pure and Applied Mathematics Volume 119 No. 16 2018, 3645-3654.
4. R. Rajkumar, V. Ganapathy, “CUSTOMIZED E-LEARNING PLATFORM: FOR MAKING TEACHING AND LEARNING
MOBILE BASED APPLICATION, International Journal of Pure and Applied Mathematics Volume 119 No. 16
2018, 3635-3643.
5. R. Rajkumar, V. Ganapathy, “Internet and Communication Technology based Education Using Virtual Reality,”
Jour of Advanced Research in Dynamical & Control Systems, Vol. 10, 05, 2018.
6. R. Rajkumar, V. Ganapathy, “DETECTION OF PANIC AND RECOVERY FROM PANIC USING BRAIN COMPUTER
INTERFACE”, International Journal of Pure and Applied Mathematics Volume 119 No. 16 2018, 3655-3662.
7. R. Rajkumar, V. Ganapathy, “IOT Enabled Education Technologies to Empower Education of Things”, Journal of
Advanced Research in Dynamical & Control Systems, Vol. 10, 05, 2018.

For Communication
R. Rajkumar
Department of Computer Science and Engineering
School of Computing
SRM Institute of Science and Technology
Kattankulathur, Chennai, India.
rajkumar2@srmist.edu.in
Phone: +91 9894808403
1. It has a significant role to play in future Genetic
Engineering fields and neuro science.
2. Robots with Artificial intelligence that acts closely to
Human
3. Improving physical methods for gathering EEGs
4. Improving generalization
5. Improving knowledge of how to interpret waves
Work in Progress

Machine learning in human computer interface

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Machine learning in human computer interface

Similar to Machine learning in human computer interface (20)

More from Rajkumar R

More from Rajkumar R (13)

Recently uploaded

Recently uploaded (20)

Machine learning in human computer interface