INTRODUCTION

Drivers drowsiness has been implicated as a causal factor in

many accidents.

The NSF reported that:

51% of adult drivers driven a vehicle while feeling drowsy.
17% had actually fallen asleep.

So, real-time drowsiness monitoring is important to avoid traffic

accidents.

Previous approaches for drowsiness detection are based on :

Physical changes.
Measuring physiological changes of drivers.
A real time wireless EEG based BCI system for drowsiness detection
is proposed.

There are some studies regarding the portable BCI devices.

The portable BCI systems are usually large.

They have to transmit the EEG signal to a back-end personal

computer to process the EEG signal.

Therefore a novel BCI system is developed.

It has the advantage of small volume and low-power consumption.

It is suitable for practical driving applications.

SYSTEM ARCHITECTURE.

Fig: EEG based BCI system

 Architecture Explanation.
First, the EEG signal from the EEG electrode is amplified and filtered by
the EEG amplifier and acquisition unit in the physiological acquisition
module.

Next, the EEG signal will be preprocessed by the microprocessor unit and
transmitted to the embedded signal-processing module via a wireless
transmission unit.

After receiving the EEG signal, It will be monitored and analyzed by our
drowsiness detection algorithm implemented in an embedded signal
processing unit.

If the drowsy state of the driver is detected, a warning tone device unit
will be triggered to alarm the driver.
 WIRELESS PHYSIOLOGICAL SIGNAL
ACQUISITION MODULE.
It consists of:
EEG amplifier .
Acquisition unit.
Microprocessor unit.
Wireless transmission unit.
Size is 4cm x 2.5cm x 0.6cm.
Photographs of (a)Wireless physiological signal
acquisition module and (b)EEG headband
embedded with this module.
Photographs of (a)Embedded Signal Processing
Module and (b)Warning Tone Device.
 REAL TIME DROWSINESS DETECTION
ALGORITHM.
The flowchart of the real time drowsiness detection
algorithm includes:
Preprocessing of EEG signal.

Construction of alertness model.

Computation of deviation of EEG spectra from the alert

model.

Implementation of the real time drowsiness detection

algorithm.
Flowchart Of Real-Time Drowsiness Detection Algorithm.
Build Alert Calculate Deviation
Model. from Alert Model.

Build alert
model of
Extract Alpha rhythm
EEG
spectra in Calculate
Alpha MDA.
rhythm.
R Pre O
A pr Thresh U
w T
oce Calculate old for
E ssi
E MDC. Drowsi P
G ng. ness. U
T
Extract
EEG Calculate
spectra in MDT.
Theta Build alert
rhythm. model of
Theta
rhythm.
 Preprocessing Of EEG Signal.
Three EEG electrodes are used to monitor the EEG signal in the
occipital midline.
The input electrode is placed in the occipital midline
The reference electrode is placed behind the right ear
The ground electrode is placed in the center of the forehead.
 Procedure of EEG Preprocessing.
Extract
EEG
spectra in
Alpha
rhythm.
Low-Pass Down
filter with sample to 512-Pt FFT
Raw
Cutoff sampling with overlap
EEG.
frequency of rate of 448-Pt
32Hz 64Hz

Extract
EEG
spectra in
Theta
rhythm.
 Construction Of The Alertness Model.
3-min EEG spectral data are used to derive the alert model.

It can be represented by a multivariate normal distribution

N(,), where is the mean vector and is the variance-co-
variance matrix.

Using Mardias test, we check whether the EEG spectra in and

rhythms follow multivariate normal.

If the alert model passes Mardias test, the alert model will be
accepted.

Otherwise, the next 3 min of EEG data will be used to derive and
validate the alert model using Mardias test again.
 Computation Of Deviation Of EEG Spectra

The alert models of alpha and theta rhythms are represented by

( ,)A and ( ,)T .

Deviation from the alert model can be calculated by:

T -1
MDA(XA)=(XA-) () (XA-).
T -1
MDT(XT)=(XT-) () (XT-).
A linear combination MDC of MDT and MDA is used to
compute a combined measure of deviation as:

MDC= X MDA + (1-) X MDT, 0 1.

MDA, MDT, and MDC can be taken as indicator of

drowsiness.

Finally, the threshold of Mahalanobis distance for

drowsiness can be defined.

If the value of MDC is larger than the threshold, the

cognitive state of the driver can be viewed as drowsy state.
Flowchart of the real time drowsiness detection algorithm
implemented in the embedded signal processing module.
Create Joint
Thread. Thread.
Thread 1:Receive EEG data
from Bluetooth.

Thread 2:EEG Pre-processing.

Thread 3:512-Pt FFT with

overlap 448-Pt.

Thread 4:Build alert model.

Thread 5:Calculate
MDA,MDT,MDC.
Thread 6:Trigger warning
tone.
 Comparison Between Our System And Other
BCI System.
BCI System Farshichi The Proposed System

Fetched EEG EEG

Signal.
Channels 6 1

EEG Cap Helmet Headband

Model.
Transmission Zigbee Bluetooth
Operation 3V 3.7V Li Battery
Voltage
ADC 8 12
Resolution(Bits)
Sampling Rate 1-100 512
(Hz)
Gain 200 5040
Frontend Signal Atmel
MSP430
Processing MCU
Unit
Backend Signal Personal ADSP-BF533
Processing Unit Computer
 ADVANTAGES.
Low power consumption and small volume.

Feasible for further extension.

High gain.

Our BCI system uses only 3 electrodes.

The setup of our BCI system is relatively easier.

 APPLICATIONS.
Our BCI system is suitable for:
Car driving applications.
Industrial Applications.
Medical Applications.
Military Applications.
 REFERENCES
J. A. Horne and L. A. Reyner, Sleep related vehicle
accidents, Brit.Med. J., vol. 310, pp. 565567, 1995.

G. Maycock, Sleepiness and driving: The experience of UK

car drivers, J. Sleep Res., vol. 5, pp. 229237, 1996.

G. Maycock, Sleepiness and driving: The experience of heavy

goods vehicle drivers in the UK, J. Sleep Res., vol. 6, pp.
238244, 1997.
 DRESTIEN TEAM
BALLARI INSTITUTE OF TECHNOLOGY &
MANAGEMENT