IEEE Consumer Electronics Magazine - October 2015 - 115

Besides, it is important to monitor the psychophysiological
conditions [87] to define whether the driver is drowsy. Some
of the methods proposed to identify the opening and closing
of eyelids are the following. Markov fields are used to
accomplish the correct iris identification [30]. Fuzzy logic is
used as a classifier to determine facial features, even determine facial expressions [88]. Diverse types of neural networks could be used for this matter. It would be preferable to
use the ones that are focused on several features [87]. A
combination of fuzzy-neural models is used usually to determine the drowsiness degree, instead of the opening and closing of eyelids [47].
The sequence followed by a vision-based system that monitors drivers' physiological features continuing the process (see
Figure 3) is shown in Figure 4. Each stage has an input and an
output, which are usually images and/or data. The stage of
informing the user about a possible danger because of drowsiness should have an action as an output. The term action
means an audible/visual alert or a vibration stimulus.

information and sensors to measure the driver's steering
behavior are combined to determine the driver's drowsiness
in real time. Another method combining multiwavelet frequency band energy with support vector machines to determine the driving state is documented in [90].

LANE DEPARTURE

HRV

Lane departure is also known as lane position [8]. The DFD systems using this driving reaction are also known as lane-departure warning systems [8], [41], [51]. The main importance of
this type of detection is based on the fact that the drowsy or inattentive user usually drifts when driving. The objective of these
DFD systems is to warn the user whenever there is an unintentional lane departure. Most of these DFD systems are based on
image recognition and rely on painted roadway markings, but
they can also be based on radar or global positioning systems.
Some reported systems based on lane departure as the primary
DFD measure are Vision/Radar Sensor (Mobileye NV) [8],
HaulCheck (AcuMine), SafeTrac (AssistWare Technologies),
AutoVue (Iteris, Inc.) [8], [41], Adaptive Lane Departure Warning System (Royal Institute of Technology sponsored by Scania,
Sweden) [41], and Lane Departure Warning (General Motors).
The General Motors system was shown at the 2007 New York
International Auto Show, and it was displayed in a 2008 Cadillac
STS. Lane Departure Prevention and Lane Departure Warning
by Nissan was also presented at the 2007 New York International Auto Show as part of the Infiniti FX and M vehicles
(Nissan's luxury line).

Another physiological element is the HRV, which is measured
by means of an electrocardiogram or arterial pressure tracing.
It can also evaluate the workload and emotional tension in an
individual [47]. Denso assembled a demonstration at the 2007
Detroit Auto Show. The system monitors eye features by
means of a camera and processes the HRV of the driver, which
is sensed through the steering wheel. The intrusive aspect is
diminished considerably by sensing the HRV through the
steering wheel [39]. Aichi Prefectual University (Japan) used a
neural network to gather and process all possible data obtained
by measuring HRV waves and steering-wheel data. There were
three indexes that can be derived from the HRV data for this
case: sympathetic nerve activity, parasympathetic nerve activity, and HRV [15]. Kanazawa University (Japan) along with
Honda studied how the beat-by-beat BP increased gradually
when driving in a monotonous environment, and this stage was

sTEERING-WHEEL MOVEMENT
Steering-wheel movement is a driving reaction, like lane
departure. It could be used as a nonintrusive method and
avoids the inclusion of a camera in the vehicle. The knowledge of the steering-wheel movements implies measuring the
angle and velocity of the steering wheel. Some reported
works are steering-wheel-angle detection using an artificial
neural network (George Washington University) [9] and
ASTiD (Pernix Ltd.), which has been used previously in the
mining sector with very positive results [8]. The ASTiD system uses a combination of a user's previous input data and
the steering-wheel behavior information. A fusion data method is reported in [89]. In this work, fusion of camera

The DFD systems based on eye
features involve the following
stages: facial extraction methods,
face detection and tracking,
eye-area detection and tracking,
upper and lower eyelid detection,
and eye feature measurements.

Digital
Image

Digital
Image Data

Eye Detection
and Tracking

Data

Physiological
Feature
Monitoring

Define
Drowsiness
State

Data

Make Driver
Aware of Current
Condition
Action
FIGURE 4. The general scheme for a DFD vision system.

october 2015

IEEE Consumer Electronics Magazine

115

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - October 2015