SUBMITTED TO: SUBMITTED BY:

MRS. JAIMALA GAMBHIR ANKIT BALODA (11104006)

ASHISH AGGARWAL (11104010)
AYESHA GUPTA (11104011)
AYUSHI GARG (11104012)
BALPREET SINGH (11104013)
Research Objective

EEG Based Drowsiness

Detection Using Mobile
Device for Intelligent
Vehicular System
Significance of the
Problem
It is estimated that 100,000
drowsiness/fatigue related Crashes Annually
Fatigue is the most frequent contributor to
crashes. A drowsy/sleepy driver is unable to
determine when he/she will have an
uncontrolled sleep onset
An accident involving driver drowsiness has a
high fatality rate because the perception,
recognition, and vehicle control abilities
reduces sharply while falling asleep
Driver Drowsiness Detection
Technique
1. In this system driver has to wear USB EEG headset
which is connected to mobile device.

2. EEG will capture live brain signals and send to

mobile device.

3. Mobile device has a application installed in it which

classifies these live signals and capture sleep
/drowsiness related signals.

4. Also, if mobile device finds a drowsiness signals it

will activate alert in it this enables the driver to wake
up or be alert.
Approach for Drowsiness
Detection and Driver
Warning
EEG dataset Collection
Alert Module
REQUIREMENTS APPLICATIONS
Hardware Sleep and drowsiness
EEG Headset (real time
implementation) ,
Mobile Device with 4-16
Gb RAM per processor
and 64-bit processors
Software
Windows Operating
System EEGLAB v12 -
MobiLaB toolkit,
MATLAB 7.6 or later
signal processing tool
box
detection
Sleep disorders like
insomnia detection
Studying brain activities of
coma patients and to
diagnose many other
conditions which affect
the brain.
This study discusses about
how EEG can be used to
implement drowsiness
detection in intelligent
transportation system for
example, cars, airplanes,
helicopters etc. to monitor
drowsiness status of the
driver /pilot
NEUROPHONE: BRAIN-
MOBILE PHONE INTERFACE
USING A WIRELESS EEG
HEADSET
 In this paper, they have proposed to use neural
signals to control mobile phones for hands-free,
silent and effortless human-mobile interaction.

Until recently, devices for detecting neural

signals have been costly, bulky and fragile.

Here they present the design, implementation

and evaluation of the NeuroPhone system,
which allows neural signals to drive mobile
phone applications on the iPhone using cheap
off-the-shelf wireless electroencephalography
(EEG) headsets.
 EEG signals from the headset are transmitted
wirelessly to an iPhone, which natively runs a
lightweight classifier to discriminate P300 signals
from noise. When a persons contact photo triggers a
P300, his/her phone number is automatically dialed.

Given the recent availability of low-cost wireless

electroencephalography (EEG) headsets and
programmable mobile phones capable of running
sophisticated machine learning algorithms, we can
now interface neural signals to phones to deliver
new mobile computing paradigmsusers on-the-go
can simply think their way through all of their
mobile applications.
P300
TheP300(P3) wave is anevent related
potential(ERP) component elicited in the process of
decision making. It is considered to be an
endogenous potential, as its occurrence links not to
the physical attributes of a stimulus, but to a
person's reaction to it. More specifically, the P300 is
thought to reflect processes involved in stimulus
evaluation or categorization.

When recorded byelectroencephalography(EEG), it

surfaces as a positive deflection in voltage with a
latency (delay between stimulus and response) of
roughly 250 to 500 ms. The signal is typically
measured most strongly by the electrodes covering
theparietal lobe.
P300
When somebody concentrates on a task-
specific stimulus (e.g., a highlighted image in
Dial Tim) among a pool of stimuli (e.g., non
highlighted images), the task-related stimulus
will elicit a positive peak with a latency of
about 300ms from the stimulus onset in
subjects EEG signal. This positive peak is
known as the P300 signal in neuroscience
literature . P300 is emanated in the central-
parietal region of the brain and can be found
more or less throughout the EEG on a number
of channels.
 We demonstrate a brain-controlled address-book
dialing app, which works on similar principles to a
P300-speller brain-computer interface: the phone
flashes a sequence of photos of contacts from the
address book and a P300 brain potential is elicited
when the flashed photo matches the person whom
the user wishes to dial.

They also demonstrate a version of the same app

which detects the much larger and more easily
detectable EEG signals triggered by the user
winking their eyes when the target photo appears.

This wink-triggered dialing works robustly in

noisy conditions. The P300, or think-triggered,
dialer is very promising but at present less reliable.
P300 Speller
 One could argue that other hands off types
of actuation such as voice recognition is
more suitable an interface to mobile
applications.

However, the goal is to best understand how

firing neurons can drive mobile applications
and what the current limitations in the state
of the art are when using off-the-shelf
wireless EEG headsets and phones.
SITUATIONS WHERE IT COULD
BE USED
We also imagine new many-to one mobile
applications; for example, a teacher of a foreign
language is interested in seeing exactly how
many students actually understood the last
question she asked. The students are all wearing
EEG headsets and their data is being streamed
in real-time to the teachers mobile phone. She
simply takes out her mobile phone and it gives
her up to the second statistics on each of her
students. She quickly glances at the aggregate
class statistics and realizing that the students
really did understand her difficult question,
proceeds with her lecture.
 In addition, the Emotiv headset, which is designed primarily
for gaming purposes, is also capable of detecting certain facial
expressions (e.g., smile, laugh, shock eyebrows raised, anger
eyebrows furrowed) and non-conscious emotions. If one
could read the emotional state of people moving through a
building then the notion of mood music would take on a literal
sense.
LIMITATIONS
Research-grade EEG headsets are expensive (e.g., tens of thousands of dollars)
but offer a much more robust signal than the cheaper (e.g., $200-$500) headsets.

As a result there is a significant amount of noise in the data of the cheaper

headsets, requiring more sophisticated signal processing and machine
learning techniques to classify neural events (e.g., P300).

The cheaper headsets provide an encrypted wireless interface between the

headset and computer allowing for mobility but complicating the design of a
clean brain-mobile phone interface.

Mobile phones are not designed to support continuous neural sensing

applications. The energy cost of continuously streaming raw neural signals over
wireless and running classifiers on the phone is challenging.

Filtering out components of the signal associated with artifacts (e.g., neural
signals associated with walking or unintentional facial expressions) is needed to
advance this vision.
 As this vision gathers speed and noise issues are
solved, EEG will be integrated into wearable
fabric (e.g., baseball caps, woolen hats, bicycle
helmets) or become the new wireless earphones
plus (i.e., earphones plus a limited set of
electrodes).

This raises a number of interesting issues. For

example, the NeuroPhone system relay transmits
raw unencrypted neural signals over-the-air to
the iPhone in IP packets. This leads to the notion
of insecure neural packets everywhere,
opening up important privacy challenges that
need to be addressed.
The Dial Tim application works on similar principles to P300-speller brain-
computer interfaces: the phone flashes a sequence of photos of contacts from the
address book and a P300 neural signal is elicited when the flashed photo
matches the person whom the user wishes to dial. EEG signals from the headset
are transmitted wirelessly to an iPhone, which natively runs a simple classifier to
discriminate P300 signals from noise. When a persons contact-photo triggers a
P300, their phone number is automatically dialed. In this case, the user wants to
dial Tim, thus when his picture is flashed, Tim is automatically dialed.
 Wireless EEG Headset
We use the Emotiv EPOC headset which has 14
data collecting electrodes and 2 reference
electrodes. The electrodes are placed in roughly
the international 10-20 system and are labeled as
such. The headset transmits encrypted data
wirelessly to a Windows based machine; the
wireless chip is proprietary and operates in the
same frequency range as 802.11 (2.4Ghz).
Also built in the headset is a gyroscope that
detects the change of orientation of subjects
head. The headset is not meant to be an
extremely reliable device, thus it is challenging to
extract finer P300 signals from the EEGs this
headset produces.
But this headset can be easily deployed at large
scale because of its low price, and can be
extremely handy if we can extract useful signals
(e.g., P300) from it through smart signal
processing and classification algorithms running
on the phone.
Design Considerations

Signal to Noise Ratio (SNR): Since the Emotiv headset is

not intended for finer signal detection, there is more noise
than usual on every electrode of the EEG. A sensible
approach to increase the SNR is to average the data over
many trials, which is also a commonly used technique in
neuroscience.

Signal Processing: Although we are averaging data for a

better SNR, we can still improve the EEG signals for better
P300 detection. We use a bandpass filter to get rid of any
noise that is not in the P300 frequency range. Again this
signal processing is unnecessary for wink mode because
wink signals are much more easily detectable in raw EEG
data.
 Phone Classifiers: Typically, real time EEG signal
processing and classification algorithms are designed for
powerful machines, not resource limited mobile phones.
These classification algorithms are not practical to run on
the mobile phone because of power efficiency and resource
issues. To address this challenge, we combine two
approaches for efficient classification on the phone:
i) we do not supply all channels from the headset to the
phone for classification, rather, only the relevant subset of
EEG channels; and ii) we implement lightweight classifiers,
more specifically, a multivariate equal-prior Bayesian
classifier is used for wink mode and a simple decision stump
is used for the think mode.
 This classifier is then applied to the rest of the
data to test whether it can generalize to unseen
data by calculating the classification precision
(i.e., percentage of classified winks that are
actually real winks), recall (i.e., percentage of real
winks that are actually classified as winks) and
accuracy (i.e., percentage of all events that are
correctly classified). The experiment results are
shown in Table 1.
 For think mode, we test on the same set of subjects. We carry
out the P300 experiments with the subjects using the Dial Tim
application while sitting still, sitting with loud background
music, and standing up. We average the data over a set time
interval. The accuracy values of the experiments are shown in
Table 2.
BRAIN RESPONSE TO
MARKETING STIMULI
USING EEG
 Background
Application of neuroscience methods to analyze and understand
human behavior related to markets and marketing exchange
has recently gained research attention.

The basic aim is to guide design and presentation of products to

optimize them to be as compatible as possible with consumer
preferences.

The general assumption is that human brain activity can provide

marketers with information not obtainable via conventional
marketing research methods (e.g., interviews, questionnaires,
focus groups). This is mainly driven by the fact that people
cannot (or do not want to) fully explain their preferences when
explicitly asked; as human behavior can be (and is) driven by
processes operating below the level of conscious awareness.
The main goals in such neuromarketing research are first to detect the small
changes in commercial stimuli that may prove to have substantial impacts on
marketing efficacy). Secondly, it also aims to explain how changes in the
depiction or presentation of marketing information affect the ways in which
the brain reacts (changes in the brain signals).

Objectives
To observe and evaluate the cortical activity of the different
brain regions and the interdependencies among the
Electroencephalogram (EEG) signals from these regions

Unlike most research in this area that has focused mainly on

liking/disliking certain products, here a way is provided to
quantify the importance of different cracker features that
contribute to the product design based on mutual
information. Here Emotiv EPOC wireless EEG headset and
Tobii-Studio eye tracker system were used to relate the EEG
data to the specific choice options (crackers).
Analysis
The commercial Emotiv EPOC wireless EEG headset with
14 channels to collect EEG signals from participants along
with a Tobii-Studio eye tracker system were used to relate
the EEG data to the specific choice options (crackers).

Subjects were shown 57 choice sets;

each choice set described three
choice options (crackers).

The patterns of cortical activity were

obtained in the five principal frequency
bands, Delta (04 Hz), Theta (37 Hz),
Alpha (812 Hz), Beta (1330 Hz), and
Gamma (3040 Hz).

Experimental Setup
A sequence of 57
choice sets was
developed. Each
described three
crackers that varied
in shape, flavor and
topping.
Emotiv EPOC
The Emotiv EPOC is a high resolution, neuro-signal
acquisition and processing wireless headset that monitors
14 channels of EEG data and has a gyroscope measure for 2
dimensional control. The electrodes are located at the
positions AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4,
F8, AF4 according to the International 1020 system
forming 7 sets of symmetric channels.
The EPOC internally samples at a frequency of 2048 Hz,
which then gets down-sampled to 128 Hz sampling
frequency per channel, and sends the data to a computer
via Bluetooth.
Both of the EPOC and eye tracker were forced to start at the
same time by means of synchronization software written in Visual
Basic to start both modules together. After the data collection
step, all collected data were transferred to MATLAB for further
processing.

Detecting and removing artifacts in the EEG data due to muscle

activity, eye blinks, electrical noise, etc., is an important problem
in EEG signal processing research - Independent Component
Analysis (ICA) and discrete wavelet transform (DWT) based
denoising used to clean the EEG signals collected by the EPOC
headset.

An initial preprocessing starts with a baseline removal, or

detrending section due to the included DC offset in the EPOC EEG
readings. This is followed by a filtering step that seeks to include
only the relevant frequencies in our analysis, remove the effect
of 50 Hz noise and eliminate artifacts related to higher
frequencies.
Findings and
results
A number of features (or variables) describing the change in
the EEG power spectrum of d, h, a, b, and c were extracted
from each of the available 14 EEG sensors from the EPOC
headset.
The most relevant brain regions associated with the choice
task by estimating the mutual dependence between the
extracted features from each sensors and the
corresponding class label of preferences as indicated by
each user (in terms of shape, flavor, and topping).

This was done using the Shannons criteria and assigning

different probabilities
The PLV results to important
suggest few differentfindings,
crackers choices.
including that the frontal
channels (AF3AF4 and F3F4) and occipital channel (O1O2) were the
most synchronized channels, which in turn indicates the importance of
cognitive processing taking place at these brain regions.
The results show the importance of all of the theta, alpha, and beta that
reflected the highest PLV at the aforementioned frontal and occipital regions.
These EEG bands and the corresponding regions with the highest PLV values
was found to be very relevant for tasks involved with emotional processing of
preferred vs. non preferred marketing stimuli when these regions were studied
separately in prior work