Performance Speciﬁcations of Market Physiological Sensors for Efﬁcient

Driver Drowsiness Detection System

Messaoud Doudou and Abdelmadjid Bouabdallah

Lab. Heudiasyc, UMR CNRS 7253, Universit

e de Technologie de Compi

egne, France

Keywords:

Driver Fatigue, Drowsiness Detection, Measurement, Sensors, Physiological Signals.

Abstract:

Signiﬁcant advances in bio-sensors technologies hold promise to monitor human physiological signals in real

time. In the context of driving safety, such devices are knowing notable research investigations to objectively

detect early stages of driver drowsiness that impair driving performance under various conditions. Seeking

for low-cost, compact yet reliable sensing technology that can provide a solution to drowsy state problem is

challenging. The contribution of this paper is to study fundamental performance speciﬁcations required for

the design of efﬁcient physiological signals based driver drowsiness detection systems. Existing measurement

products are then accessed and ranked following the discussed performance speciﬁcations. The ﬁnding of

this work is to provide a tool to facilitate making the appropriate hardware choice to implement efﬁcient yet

low-cost drowsiness detection system using existing market physiological sensors products.

1 INTRODUCTION

Till now, the total number of serious car crashes is

still increasing regardless of improvements in road

and vehicle design for driver safety. Reduced men-

tal alertness due to drowsy state have been identiﬁed

as the greatest safety danger and the major cause of

road trafﬁc accidents (Lal and Craig, 2001). While

each day in the United States 80,000 individuals fall

asleep behind the steering wheel (American Academy

of Sleep Medicine, 2005), 25-30% of driving ac-

cidents in the UK are drowsiness related (ROSPA,

2001), about 35% drivers in the Netherlands and 70%

drivers in Spain have reported falling asleep while dri-

ving (Morales et al., 2015).

The measure of human physiological parameters

allows evaluating objectively cognitive-attentive indi-

cators, in reaction to external perceptual stimuli. The

study of human physiology has showed that mono-

tone driving task and nocturnal driving mostly lead

to sleep deprivation, lacking sleep, and being in a

state of low energy (Morales et al., 2015). These

symptoms decrease cognitive abilities and make dri-

ver more prone to fatal errors. Many drowsiness mea-

surement technologies have been developed to moni-

tor driving behavior and alert drivers when drowsy.

Recently, with the remarkable advance in sensing

and communication technologies, Low-cost wearable

devices are fast becoming a key instrument on bio-

sensors based applications and they have been app-

lied in many ﬁelds including industrial, transporta-

tion, medical, daily-life, sport, etc. There are a num-

ber of tentative promoted by shift-work industries to

monitor cognitive state of human-being using these

emerging technologies since they hold the promise of

being objective compared to other measuring techno-

logies. These bio-signals based technologies make it

possible to alert driver at earlier stages of drowsiness

and thereby prevent many drastic accidents providing

a solution to the driver drowsy problem (Sahayadhas

et al., 2012).

In this study, we focus in assessing recent develop-

ment of bio-sensors technologies in the market that

are currently underway to address driver drowsiness

issue, and provide a concise hardware speciﬁcation

tool for the design of efﬁcient driver alertness mo-

nitoring system. In the following, the key drowsi-

ness detection technologies are presented in section

2. Next, the general architecture of driver drowsiness

monitoring system with different modules are explai-

ned in section 3. Main performance characteristics

that must be met by a drowsiness monitoring techno-

logy are discussed in section 4. Section 5 is devoted

to review potential market physiological sensors pro-

ducts. Ranking methodology is described in section

6 providing a tool to make the appropriate hardware

choice of existing products. Finally, section 7 conclu-

des the paper.

Doudou, M. and Bouabdallah, A.

Performance Speciﬁcations of Market Physiological Sensors for Efﬁcient Driver Drowsiness Detection System.

DOI: 10.5220/0006607800990106

In Proceedings of the 7th International Conference on Sensor Networks (SENSORNETS 2018), pages 99-106

ISBN: 978-989-758-284-4

2 DROWSINESS DETECTION

TECHNOLOGIES

A plethora of driver fatigue researches exist span-

ning different measurement technologies. The most

commonly used measurement can be categorized

upon the monitoring instrument into: (i) Vehicle-

based sensors, (ii) Video-based sensors, and (iii)

physiological signals sensors such as electrooculo-

graphy (EOG), electromyography (EMG), electrocar-

diography (ECG), and electroencephalography (EEG)

signals where the latter is the most used (Sahayadhas

et al., 2012; Sanjaya et al., 2016).

Vehicle monitoring focuses on driving and vehi-

cle patterns such as steering wheel angles and rever-

sals, the car position with respect the road’s middle

lane and the standard deviation of lane position

(SDLP), etc. However, this technology can operate

reliably only at particular environments (Sahayadhas

et al., 2012) depending on the geometric characteris-

tics of the road and to a lesser extent on the kinetic

characteristics of the vehicle (Vural, 2009) and they

are easily inﬂuenced by other factors such as road

marking, climate, lighting and trafﬁc conditions.

Video monitoring measures driver’s facial ex-

pression and detect drowsiness from differentiating its

abnormal behavior such as eye closure (PERCLOS),

head nodding, etc. The common limitation is lig-

hting. The detection rate using this technology was

59% compared to 85% and 97.5% of EEG and ECG

(Sanjaya et al., 2016). (Golz et al., 2010) reported

an accuracy around 74% and 66% using PERCLOS

compared to 87% and 89% from EEG/EOG.

Bio-sensors measure physiological signals from

organs such as brain, eyes, muscles, and heart which

have visual correlation with fatigues/drowsiness. Re-

searchers observed via EEG that drivers had sleep

bursts accompanied by theta waves and K-complexes

while they still had their eyes open, something video-

based monitoring might have missed. Physiological

signals have been shown to be reliable and accurate

since they are less impacted by environmental and

road conditions and thus may have fewer false positi-

ves than other measures (Zilberg et al., 2007).

3 SYSTEM ARCHITECTURE

Due to the increasing interest in the use of wearable

bio-sensor systems, many communication architectu-

res have been proposed depending on the target ap-

plication (Lee et al., 2013). The general architecture

of bio-sensor system is composed by three main mo-

dules: (i) signal acquisition, (ii) data processing, and

Warning/Alert Buzz Speed Control

Control Module

Wirless/Wire Transfert

Translation &

Feature Extraction

Training/Classiﬁcation

Algorithms

Processing Module

Wirless/Wire Transfert

BioSensors (EEG,

EOG, etc.)

Noise/Artifact

Removal

Acquisition Module

Figure 1: Logical view of different modules of the system.

(iii) control modules as depicted in Fig. 1.

3.1 Acquisition Module

This module is composed of different physiological

wearable sensors such as EMG, ECG, EOG, EEG,

etc. attached to the body which measure physiolo-

gical signals. These sensors form a network and com-

municate with the network coordinator to send data.

The measured signals are then ﬁltered and transfor-

med to remove any noise and artifact that may affect

the quality of sensed data values.

3.2 Processing Module

Signals are received from acquisition module after ﬁl-

tering noise and removing artifacts. As second stage,

signals are processed to extract the main features that

reﬂect different states of the target application (e.g.

the cognitive states of driver). These features are then

passed to the training and classiﬁcation algorithms

to determine the new measured states. As for driver

drowsiness, the features can be used to determine in

which level of alertness the driver is.

3.3 Control Module

Driver alertness is monitored in real time using acqui-

sition and data processing modules. Whenever a

drowsy state is identiﬁed, the detection event is then

triggered by the control module to make the appro-

priate action in time. This action may be an alarm or

buzz inside the vehicle to alert or wake-up the driver.

The action may take control of the vehicle in order to

speed-down or stop the vehicle.

Many portable systems propose to incorporate the

acquisition and the processing modules into the same

component to compact the system. Hence, there is a

SENSORNETS 2018 - 7th International Conference on Sensor Networks

100

Figure 2: Driver Drowsiness Detection System Architec-

ture.

serious issue with the battery lifetime. In the context

of driver drowsiness detection, the acquisition module

is attached to the driver and the processing module

is installed on the vehicle which has sufﬁcient power

supply. This allows extending the battery lifetime and

keep monitoring for long periods. The control module

is mounted on the vehicle to trigger warning messages

and sound alerts. This module can be even enabled to

control some actions of the vehicle such as accelera-

tion and speed. The system can be extended to sup-

port multi-tiers cloud-based architecture (Zao et al.,

2014). As depicted in Fig. 2, some of data can be

transmitted via 3G/4G/LTE connections to the remote

servers where data analytic algorithms can be used to

train and extract new knowledge. This enables moni-

toring cognitive states during real driving tasks from

large number of drivers and may be explored by the

research community to enrich training sets and im-

prove the accuracy of existing detection algorithms.

4 PERFORMANCE

SPECIFICATIONS

If any bio-sensors system is to prove suitable for de-

tecting driver drowsiness, it must meet some perfor-

mance speciﬁcations. These speciﬁcations are essen-

tial in making the appropriate bio-sensors hardware

choice for design consideration. In the following, the

major speciﬁcations are discussed:

4.1 Multi-sensors Support

Single signal measurement such as EEG may neces-

sitate dense electrode placement in different locations

to accurately capture cognitive states. Hybrid signal

acquisition through simultaneous recording of diffe-

rent bio-signals can yield higher accuracy of the sy-

stem. Combination of multiple bio-signals measure-

ments, such as ECG, EMG, EOG with EEG, the sy-

stem can measure not only brain waves but also heart

rate, eye movements, etc. Research results have sho-

wed that adding either EOG or ECG measurements,

there is further improvements in reduction of error ra-

tes in drowsy state detection (Warwick et al., 2015).

4.2 Type of Electrodes

The choice of electrode technology is very impor-

tant since it represent the sensing component. With

respect EEG measurement, wet electrodes known as

silver-chloride electrodes (Ag/AgCl) are widely used

by current market products. These electrodes are low-

cost, and have low contact impedance, and good sta-

bility in time. Wet electrodes requires removing outer

skin layer of the scalp and ﬁlling a special conductor

gels which take long time to prepare and are uncom-

fortable to users. Dry electrodes are other technology

which do not need to use gel and skin cleaning. Ho-

wever the bad signal quality is their main disadvan-

tage. Fig. 3 shows an example of wet and dry electro-

des available in the market.

(a) (b)

Figure 3: (a) Wet electrodes vs. (b) Dry electrodes.

4.3 Electrode Placement

Capturing as much as data from strategically locati-

ons is critical to pinpointing the drowsy related cau-

ses. For each bio-signal, there exists suitable locati-

ons where may be placed to efﬁciently measure signal

reﬂecting the drowsy state of driver. For EOG, elec-

trodes are attached to the eye skin (up/down/left/right)

whereas for EMG, they may be placed on the left

bicep, right bicep, left forearm ﬂexor, right forearm

ﬂexor, frontal muscles, or on the deltoid, trapezius

Hostens and Ramon (Hostens and Ramon, 2005).

Performance Speciﬁcations of Market Physiological Sensors for Efﬁcient Driver Drowsiness Detection System

101

While 5 & 12 lead electrode placements are gene-

rally used for ECG recording. For EEG, the electrode

placement according to the 10-20 Standard deﬁnes

which brain location that serves a speciﬁc function

(see Fig. 4). More speciﬁcally: Prefrontal Cortex (Fp)

for emotional inhibition and attention; Frontal Lobes

(F) for working memory, metaphorical thinking, sus-

tained attention and judgment; Central Strip (C) for

sensory-motor functions; Temporal Lobes (T) for lan-

guage comprehension and long-term memory; Parie-

tal Lobes (P) for language processing and procedural

memory; Occipital Lobes (O) for visual processing.

Thus, locations concerning various forms of attention

which reﬂect alertness/drowsiness states must be co-

vered by the hardware.

Figure 4: The 10-20 system of EEG electrode placement.

4.4 Number of Channels

An electrode capturing bio-signal activity is called a

channel. Typical Bio-sensor systems can have as few

as a single channel to as many as channels (256 for

EEG) depending on the required density. The sy-

stem must trade-off between capturing as much as

bio-activities with some performance metrics. For in-

stance, increasing the number of channels will have

signiﬁcant delay for data processing. Second, more

channels mean higher costs and more difﬁcult expe-

rimental setups. Lastly, by increasing the number of

channels, the huge amount of signals will be trans-

mitted that impairs reliability and battery usage es-

pecially for mobile and low-power systems. On the

other hand, very few channels impair the accuracy of

detection.

4.5 Portability & Mobility

Conventional bio-sensors systems such as actiCHamp

(Brain Products), Neuroscan NuAmps Express (Com-

pumedics Ltd.), and EDVTCS (Neurocom) are wi-

red. The acquisition part of wired systems gene-

rally comes with bulky and heavy ampliﬁers and pre-

processing units. Connecting wires is usually compli-

cated with a large number of cables between the elec-

trodes and the acquisition part. For these reasons, pre-

paration time for measuring signals is typically very

long. In addition, user movement is limited due to ca-

ble constraints. Therefore, the application of drow-

siness detection based on these systems is difﬁcult

to escape from laboratory scale experiments. With

emerging wearable technologies, biopotential measu-

rements, such as EEG, ECG, EMG, and EOG can be

delivered in real-time via wireless and Low-energy

connections such as WiFi, Bluetooth, ZigBee, etc.

Therefore, these provide the advantages of mobility

and long-term monitoring. Portable systems facilitate

the implementation of driver drowsiness detection sy-

stems and enable in-ﬁeld experimentation instead of

simulation environment. However, huge volume of

signals may be sampled and need to be transmitted

wirelessly in real-time. Hence, the system must prove

energy-efﬁcient operation for long period to be accep-

ted for continuous monitoring. for example, compres-

sion algorithms can be used to alleviate big data trans-

fer since it is time and energy consuming (Hussein

et al., 2015).

4.6 Artifact Removal

Bio-sensors are prone to various sources of noise

and artifacts. Signal conditioning is essential to ena-

ble transmission of precise bio-signals. Many noise

sources are likely present from physiological inter-

ference and power line noise. Physiological interfe-

rence occurs between EEG, EMG, ECG, EOG and

others. The amplitude of EMG, ECG and EOG is re-

latively lager around 50uV and 20-30mV while that

of EEG is much smaller around 10 100uV. Thus, the

EEG signals are easily buried by these physiological

signals unavoidably. Power line noise (Outlet, USB,

etc.) can also contaminates the EEG signals in the

range of 50 or 60Hz. Furthermore, the measured bio-

signals of mobile systems are also subject to heavy

motion and vibration artifacts. Hence, the presence of

noise and artifact removal mechanism is essential in

such systems.

4.7 System Autonomy

Another important speciﬁcation is the need of energy-

efﬁcient and long-term wearing system. For systems

that use battery powered bio-sensors, the lifetime of

the system is the critical challenge to ensure conti-

nuous driver monitoring. In fact, wireless transmis-

sions consume the largest amount of device’s energy.

Indeed, the battery autonomy may go from 4 hours

to 24 hours or even more depending on the wireless

SENSORNETS 2018 - 7th International Conference on Sensor Networks

102

Brain Products Biosemi Mind Media ANT Neuro Cognionics G.tec NeuroScan

ActiCAP ActiveTwo NeXus-32 eego sports HD-70 Nautilus QuickCaps

QUASAR - DSI 10/20 NeuroElectrics - Enobio mBrainTrain ABM OpenBCI IMEC OpenEEG

Emotiv - EPOC+ NeuroSky - MindWave Macrotellect - BrainLink InteraXon - Muse Versus Melon Focus

Figure 5: Overview of biosensors market products.

technology (e.g. Bluetooth, Wiﬁ, etc.) and on the

sampling rate. The system must be designed with ef-

ﬁcient usage of sensory and radio components to ens-

ure reasonable monitoring lifetime.

4.8 Software

The software is one of the main part of the system.

Thus it is fundamental to have access to data in or-

der to manipulate and/or analyze the recorded signals.

The market product may provide software develop-

ment kit (SDK) as bio-signal acquisition software or

an application programming interface (APIs) compa-

tible with some known commercial or open source

bio-signals software platforms (e.g. BCI 2000, Open-

Vibe, LabVIEW, etc). This enables to facilitate and

speed-up the development of efﬁcient detection algo-

rithms.

4.9 Product Cost

Making a choice between products must trade-off sy-

stem performance with its cost. Many of existing

market devices are designed for clinical and research

purpose and provide multi-sensor acquisition with a

large number of electrodes/channels and with incre-

dible sensitivity. The cost of such systems is visibly

high due to the full provided functionality. Depen-

ding on the application need like driver drowsy de-

tection, the system cost may be reduced and can be

determined by the performance speciﬁcations such as

number of channels, sensors’ type, portability, wire-

less technology, ﬂexibility, and comfort.

5 MARKET PRODUCTS

With the growing progress in sensing technologies,

more smart, compact and user-friendly products have

been increasingly introduced to the market. Each

of existing bio-sensor products provides speciﬁc mo-

nitoring functionality of human physiological states.

For example, MySignals

provides e-Health platform

that includes several bio-signals measurements such

as ECG, EMG and X-Y-Z Accelerometer but not EEG

and EOG.

ActiCHamp cap from Brain Products is destined

for EEG signal acquisition. ActiCHamp exists with

different channel and sampling rate conﬁgurations

ranging from 32 to 160 channels and from 10 to 100

kHz receptively. ActiCAP express is light head cap

version with 16 channels with active electrodes. Bio-

semi developed Active Two which is a 8/16/32 chan-

nels acquisition cap system with wet electrodes. The

eego/rt sports from ANT Neuro is a portable head

cap with up to 64 channels for rehabilitation mental

states studies, and can work without conductive gel

electrodes. NeXus-32 and Nexus-4 (Mind Media) are

32 channels (heavy) and 4 channels (portable) bio-

sensors head cap with wet electrodes. ANT Neuro

developed wireless eego/rt sports with 7-32 channels

and dry electrodes.

Cognionics developed 64 channels headset Dry

electrodes for general signals measurement. Quick-

20/30 is light version with 20 channels and pos-

sibility to integrate 8 channels from auxiliary

EOG/ECG/EMG/PPG bio-sensors. Cognionics also

designed Sleep HeadBand with 10 channels for sleep

monitoring. G.tec designed g.nautilus with 8/16/32

http://www.my-signals.com/

Performance Speciﬁcations of Market Physiological Sensors for Efﬁcient Driver Drowsiness Detection System

103

channels and wet/dry electrodes for clinical and re-

search bio-signals measurement. Quick Cap is 256

wet electrode head cap from NeuroScan capable of

measuring EEG, ECG, EMG, and EOG bio-signals.

QUASAR designed DSI-10/20 head cap with 21 dry

electrodes. While Enobio from NeuroElectrics is a

bio-signals acquisition head-cap with 8/16/32 chan-

nels. mBrainTrain designed 24 channels EEG wire-

less cap with wet electrodes as a research tool for Psy-

chology, Sport, sleep, and Serious gaming/VR stu-

dies. ABM realized B-Alert X10 (13 channels) and

B-Alert X24 (24 channels) which are portable bio-

sensors headsets that can measure EEG combined

with some other bio-signals such as ECG and provide

quick and valuable insight into the cognitive function

and mental state of the user.

OpenBCI is an open-source bio-sensors board

capable of measuring EEG, EMG, and ECG sig-

nals. OpenBCI can support 4/8/16 and wet/dry elec-

trodes which are sold separately. IMEC develo-

ped EEG headset with 8 channels to monitor Emer-

gency Room and Intensive Care Unit patients. Omi-

lex sold ModularEEG which is a 2 channels open-

source hardware known as OpenEEG. Neurosky de-

signed MindWave; a single channel EEG using one

dry electrode on the forehead (FP1) for everybody

use. Emotiv is another company that developed mo-

bile bio-sensors. EPOC+ is a 14 channels and In-

sight is a 5 channels from Emotiv that use dry elec-

trodes and are capable of providing the following me-

trics to the users: (i) Engagement/Boredom which re-

ﬂects long-term alertness and the conscious direction

of attention towards task-relevant stimuli, (ii) Excite-

ment (Arousal) that reﬂects the instantaneous arousal

towards stimuli associated with positive valence, (iii)

Stress (Frustration), and (iv) Meditation (Relaxation).

Versus is EEG headset with 5 channels and dry elec-

trodes designed for athletic peak-performance neuro-

feedback training through customized exercise pro-

tocols to improve mental acuity, concentration, and

sleep management. Muse Headset from Interaxon is

an easy-to-use 4 channels headband for concentration

and meditation training. Melon is a slim EEG he-

adset with 4 dry electrodes for focus neurofeedback.

iFocusBand is a neoprene headband with 3 ﬂexible

woven electrodes targeted primarily for sports perfor-

mance training. Table 2 provides a review of existing

market bio-sensors and highlights the main perfor-

mance speciﬁcations including the number of chan-

nels, electrode placement, data transfer technology

and sampling rates as well as the battery autonomy

and the corresponding cost whenever provided.

6 PRODUCTS RANKING &

DISCUSSION

Notable efforts are taking place to promote bio-

sensors technologies for pioneer applications. To

our knowledge, there exists practically very few bio-

sensors product intended for driver drowsiness de-

tection on the market. Most of existing products

provide bio-signals monitoring for general research

usage or for medical and neurofeadback applications

such as training, sport, gaming, etc. In the context of

driving monitoring, more efforts are needed to meet

performance speciﬁcations to develop efﬁcient drow-

siness detection system. For instance, high precision

products are bulky and rely upon a large number of

channels (e.g., 64-256), which is cost non-effective

and makes it difﬁcult to do fast artifact removal. Furt-

hermore, electrode placement is too technical due to

the requirement for electrodes, gel, wiring, etc. The

use of dry electrodes is promising to reduce the cost

and time required for data collection but novel techni-

ques are needed to improve the accuracy of measured

signals. Lower cost products come with reduced reso-

lution (e.g., 4-16 channels) but with increased porta-

bility. Although, these devices are cost effective and

more comfortable, they either suffer from low accu-

racy and require additional signal inputs such as EOG,

ECG, EMG to maintain high accuracy.

In the context of driver drowsiness detection, it

would be preferable that the bio-sensor system is less

obtrusive and composed with multiple bio-sensors es-

pecially EEG and EOG (Golz et al., 2010), with few

but sufﬁcient number of channels, active electrodes,

low-power communication technology with accepta-

ble sampling rate and battery autonomy. To facilitate

the choice of suitable hardware for drowsiness de-

tection, we have ranked the reviewed bio-sensor pro-

ducts in Sec. 5 using the performance speciﬁcations

discussed in Sec. 4. As multiple bio-sensors are nee-

ded, we ranked with (1,2,3,4) whenever EEG, ECG,

EMG, EOG are supported. Electrode type is ranked

with 1 for wet and 2 for dry. Electrode placement is

ranked only for EEG from 1 to 6 for (Fp) (F) (C) (T)

(P) (O) locations. We ranked the number of channels

with 1/2/3/4/5 for 64-256/32-64/16-32/8-16/1-8 chan-

nels. Portability is ranked following the data transfer

technology as 1/2/3/4 for USB, Wiﬁ, BLE, RF

. Arti-

fact removal is ranked with 0 or 1 for the existence of

the mechanism. Battery lifetime is ranked as 1/2/3/4/5

for 1-4/4-8/8-12/12-16/16-24 hours autonomy. The

BLE: Bluetooth Low Energy marketed as Bluetooth

Smart. RF: Proprietary RF refers to any radio frequency

speciﬁc to an original equipment manufacturer OEM and it

is under 928MHz.

SENSORNETS 2018 - 7th International Conference on Sensor Networks

104

software is ranked with 0/1/2 when the signal proces-

sing software is provided and wither is commercial

or open-source software. Finally, the cost is ranked

with 1/2/3/4/5 for price ranging in [50k 100k]/[25k

50k]/[10k 25k]/[1k 10k]/[0 1k]$.

Table 1 shows the results of existing bio-sensor

products ranking. It can be observed that Open-

BCI, Enobio, and DSI10/20 are the top ranked pro-

ducts that met the required performance speciﬁcations

among others. This ranking is based on our study of

physiological based-sensors technologies and we be-

lieve that it is not the only evaluation and ranking met-

hod to access the performance of such technologies.

Although some performance metrics were not taken

into consideration in our ranking (e.g., device com-

fort), we think that the proposed ranking tool help in

choosing the most appropriate hardware products to

develop efﬁcient drowsiness detection system.

7 CONCLUSION

Driver drowsiness poses a major danger for public

safety. Monitoring driver’s alertness is of high im-

portance to prevent grand number of incidents. Exis-

ting drowsiness detection technologies such as vehi-

cle and video-based have limited accuracy and work

well in speciﬁc conditions. Recently, a number of

portable bio-sensor devices have rapidly attracted the

research interest to circumvent the drive drowsy pro-

blem under any condition. These promising devices

can objectively capture the drowsiness state by moni-

toring physiological signals of drivers and alert them

in real-time. However, the choice of the hardware

must trade-off some performances such as signal qua-

lity and the cost. This paper discusses a number of

performance speciﬁcations required by such systems

and rank the existing market physiological sensor pro-

ducts following these speciﬁcations providing the re-

search community with a tool to make the appropriate

hardware choice for design consideration of efﬁcient

yet low-cost driver drowsiness detection. We plan to

perform experimental comparison tests between some

existent market platforms in our research agenda.

ACKNOWLEDGEMENTS

This work was part of WISSD Project carried out

in Heudiasyc Lab and was co-funded by the French

Regional Program (Hauts-de-France), and the Euro-

pean Regional Development Fund through the pro-

gram FEDER.

Table 1: Bio-Sensors Products Ranking.

Product

Number of Channels

Electrode Placement

Type of Electrode

Multi-Sensors

Portability

Artifact Removal

Battery Autonomy

Software

Cost

Total

ActiCHamp-64 2 6 1 1 1 1 2 2 2 18

ActiCHamp-32 3 6 1 1 1 1 2 2 2 19

ActiCap 4 6 2 1 1 1 2 2 3 22

eego/rt sports 2 6 1 3 1 1 2 1 2 19

Active Two-128 1 6 1 3 1 1 2 2 2 19

Active Two-64 2 6 1 3 1 1 2 2 2 20

Active Two-32 3 6 1 3 1 1 2 2 3 22

Active Two-16 4 6 1 3 1 1 2 2 3 23

Active Two-8 5 6 1 3 1 1 2 2 3 24

Cognionics-70 1 6 2 3 3 1 2 1 2 21

Cognionics-40 2 6 2 3 3 1 2 1 2 22

Cognionics-32 3 6 2 3 3 1 2 1 3 24

Cognionics-24 3 6 2 3 3 1 2 1 3 24

Quick-20 3 6 2 3 3 1 2 1 3 24

Sleep Headband 4 3 2 1 3 1 2 1 4 21

G.tec SAHARA-32 3 6 2 1 4 1 2 1 3 23

G.tec SAHARA-16 4 6 2 1 4 1 2 1 3 24

G.tec SAHARA-8 5 6 2 1 4 1 2 1 4 26

Q. DSI10/20 3 6 2 4 3 1 5 1 3 28

Enobio-8 5 6 2 4 3 1 2 1 4 28

Enobio-20 3 6 2 4 3 1 2 1 3 25

Enobio-32 3 6 2 4 3 1 2 1 3 25

B-Alert X10 4 3 1 4 3 1 2 1 4 23

B-Alert X24 3 4 1 4 3 1 2 1 3 22

Quick Caps 1 6 1 4 1 1 1 1 1 17

NeXus-32 3 6 1 4 3 1 5 1 3 27

mBrainTrain Cap 3 6 1 1 3 1 2 1 4 22

OpenBCI-4 5 4 2 3 4 1 5 2 5 31

OpenBCI-8 5 6 2 3 4 1 5 2 5 33

OpenBCI-16 4 6 2 3 4 1 5 2 5 32

IMEC 5 4 2 1 3 1 5 1 3 25

OpenEEG 5 2 2 1 1 0 1 0 5 17

Emotiv EPOC+ 4 4 1 1 4 1 3 1 5 24

Emotiv Insight 5 3 2 1 4 1 1 1 5 23

NeuroSky 5 1 2 1 4 1 2 1 5 22

BrainLink 5 1 2 1 3 1 1 0 5 19

Muse 5 3 2 1 3 1 2 2 5 24

Versus 5 2 2 1 3 1 2 0 5 21

Melon 5 1 2 1 3 1 2 0 5 20

Focus 5 1 2 1 3 1 3 0 5 21

REFERENCES

Golz, M., Sommer, D., Trutschel, U., Sirois, B., and Ed-

wards, D. (2010). Evaluation of fatigue monito-

ring technologies. Somnologie-Schlafforschung und

Schlafmedizin, 14(3):187–199.

Hostens, I. and Ramon, H. (2005). Assessment of muscle

fatigue in low level monotonous task performance du-

ring car driving. Electromyography and kinesiology,

15(3):266–274.

Hussein, R., Mohamed, A., and Alghoniemy, M. (2015).

Scalable real-time energy-efﬁcient eeg compression

scheme for wireless body area sensor network. Bio-

medical Signal Processing and Control, 19:122 – 129.

Lal, S. K. and Craig, A. (2001). A critical review of the

psychophysiology of driver fatigue. Biological Psy-

chology, 55(3):173 – 194.

Performance Speciﬁcations of Market Physiological Sensors for Efﬁcient Driver Drowsiness Detection System

105

Table 2: Review of well-known Bio-Sensors market products with major performance speciﬁcations.

Company Product Channel Electrodes EEG Placement Bio Sensors Data

Transfer

Sampling

rate

(kHz)

Battery

Auto-

nomy

System

Cost ($)

Brain Products ActiCHamp 160 Wet (Fp) (F) (C) (T) (P) (O) EEG USB 10-25 6 hr 96,500

128 Wet (Fp) (F) (C) (T) (P) (O) EEG USB 10-25 6 hr 80,000

96 Wet (Fp) (F) (C) (T) (P) (O) EEG USB 10-25 6 hr 66,200

64 Wet (Fp) (F) (C) (T) (P) (O) EEG USB 25-50 6 hr 49,900

32 Wet (Fp) (F) (C) (T) (P) (O) EEG USB 50-100 6 hr 35,600

ActiCap 16 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG USB 2-20 6 hr 11,375

ANT Neuro eego/rt sports 64+24 Wet (Fp) (F) (C) (T) (P) (O) EEG EMG EOG USB 2.048 6 hr > 25,000

Biosemi Active Two 256+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 75,000

160+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 52,000

128+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 45,000

64+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 27,000

32+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 21,000

16+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 17,000

8+7 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG USB 2-16 5 hr 13,500

Cognionics Dry Head Set 16+8 Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG BLE 0.262 6 hr ∼15,500

24+8 Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG BLE 0.262 6 hr ∼20,500

32+8 Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG BLE 0.262 6 hr 26,500

64+8 Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG BLE 0.262 6 hr 42,600

Quick-20 20+8 Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG BLE 0.262 6 hr 20,600

Sleep Headband 10 Dry (Fp) (F) (T) EEG BLE 0.262 6 hr 3,800

G.tec g.sahara/nautilus 8 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG RF 0.25/0.5 8 hr 4,500

16 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG RF 0.25/0.5 8 hr ∼9,500

32 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG RF 0.25/0.5 8 hr ≤25,000

QUASAR DSI10/20 21 Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG EOG BLE 0.24/0.9 24 hr 22,500

NeuroElectrics Enobio 8 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG EOG BLE 0.25 8 hr 4,995

20 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG EOG BLE 0.25 8 hr 14,495

32 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG EOG BLE 0.25 8 hr 24,995

ABM B-Alert X10 9+4 Wet (F) (C) (P) EEG ECG EMG EOG BLE 0.256 8 hr 9,950

B-Alert X24 20+4 Wet (F) (C) (P) (O) EEG ECG EMG EOG BLE 0.256 8 hr 19,950

NeuroScan Quick Caps 256 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG EOG USB 02/0.5 0 81,396

Mind Media NeXus-32 21 Wet (Fp) (F) (C) (T) (P) (O) EEG ECG EMG EOG BLE 2.048 20 hr 23,995

mBrainTrain EEG Cap 24 Wet (Fp) (F) (C) (T) (P) (O) EEG BLE 0.25/0.5 5 hr 6,925

OpenBCI Head Set 4 Wet/Dry (Fp) (F) (C) (T) EEG ECG EMG RF/BLE 0.20 26 hr 199+60

8 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG RF/BLE 0.25 26 hr 499+60

16 Wet/Dry (Fp) (F) (C) (T) (P) (O) EEG ECG EMG RF/BLE 0.25 26 hr 899+60

IMEC EEG Headset 8 Dry (F) (C) (T) (P) EEG BLE 22 hr 25,000

Olimex OpenEEG 2 Wet/Dry (Fp) (F) EEG USB 0.19/0.5 0 119

Emotiv EPOC+ 14 Wet (F) (T) (P) (O) EEG RF 0.128 12 hr 799

Insight 5 Dry (F) (T) (P) EEG RF 0.128 4 hr 299

NeuroSky Mind Wave 1 Dry (Fp) EEG RF 0.25 8 hr 130

Macrotellect BrainLink 1 Dry (Fp) EEG BLE 0.512 4 hr 373

InteraXon Inc. Muse 5 Dry (Fp) (P) (O) EEG BLE 0.22 5 hr 299

SensLabs Versus 5 Dry (F) (C) EEG BLE 0.25/1.28 5 hr 399

Melon Inc. Head band 1 Dry (Fp) EEG BLE 0.25 8 hr 149

Focus IFocusBan 2 Dry (Fp) EEG BLE 0.25 12 hr 500

Lee, S., Shin, Y., Woo, S., Kim, K., and Lee, H.-N.

(2013). Review of wireless brain-computer interface

systems. In BCI Systems-Recent Progress and Future

Prospects, pages 215–238. Intech.

Morales, J. M., Di Stasi, L. L., D

ıaz-Piedra, C., Morillas,

C., and Romero, S. (2015). Real-time monitoring of

biomedical signals to improve road safety. In Conf. on

Artiﬁcial Neural Networks, pages 89–97. Springer.

ROSPA (2001). Driver fatigue and road accidents: A li-

terature review and position paper. Tech. Rep. Royal

Society for the prevention of accidents, Birmingham,

U.K.

Sahayadhas, A., Sundaraj, K., and Murugappan, M. (2012).

Detecting driver drowsiness based on sensors: A re-

view. Sensors, 12(12):16937–16953.

Sanjaya, K., Lee, S., and Katsuura, T. (2016). Review on the

application of physiological and biomechanical mea-

surement methods in driving fatigue detection. Me-

chatronics, Electrical Power, and Vehicular Techno-

logy, 7(1).

Vural, E. (2009). Video based detection of driver fatigue.

Ph.D. Thesis. Sabanci University. Istanbul. Turkey.

Warwick, B., Symons, N., Chen, X., and Xiong, K. (2015).

Detecting driver drowsiness using wireless wearables.

In IEEE MASS, pages 585–588.

Zao, J. K., Gan, T.-T., You, C.-K., Chung, C.-E., Wang, Y.-

T., M

endez, S. J. R., Mullen, T., Yu, C., Kothe, C.,

Hsiao, C.-T., et al. (2014). Pervasive brain monitoring

and data sharing based on multi-tier distributed com-

puting and linked data technology. Frontiers in human

neuroscience, 8.

Zilberg, E., Xu, Z. M., Burton, D., Karrar, M., and Lal, S.

(2007). Methodology and initial analysis results for

development of non-invasive and hybrid driver drow-

siness detection systems. In IEEE AusWireless, pages

16–16.

SENSORNETS 2018 - 7th International Conference on Sensor Networks

106