Signal Processing - November 2016 - 46

acquisition and analysis of vehicle measures, facial/body
as a potential hazard, but actively anticipate other drivers'
expressions, and physiological data has not been feasible
behaviors around the stalled truck. Furthermore, SA in drivuntil recently, data fusion techniques have not yet been
ing includes an internal awareness of the driver's own state
studied much in the context of smart driver monitoring.
and behaviors. For example, a driver aware of her drowsiness
As vehicles are becoming increasingly automated, it is
may decide to take a break from driving. SA may be best
becoming important for the vehicle to be able to monitor
supported through real-time feedback, helping drivers track
the driver's state to safely take control from the driver and
and understand critical elements on the road. A smart driver
transfer it back to the driver when there is
monitoring system that detects low arousa need. (For a comprehensive discussion
al state of the driver may also encourage
As vehicles are becoming
of various types and levels of automation,
drivers to take a rest break or seek out
increasingly automated, it refer to [59] and the Preliminary Statesome stimulation, such as music.
is becoming important for ment of Policy by the National Highway
A high level of SA can be demonstrated
the vehicle to be able to
Traffic Safety Administration, available
through "anticipatory driving," a cognitive
at www.nhtsa.gov/staticfiles/rulemaking/
competence in identifying traffic situamonitor the driver's state
tions through perception of cues, such that
to safely take control from pdf/Automated_Vehicles_Policy.pdf.) In
the former scenario, i.e., the transfer of
the vehicle can be positioned efficiently
the driver and transfer it
control from the driver to the vehicle,
for potential upcoming changes in traffic
back to the driver when
driver state monitoring can play a cru[53]. Feedback intended to support anticithere is a need.
cial role in the timely and accurate enpatory driving can focus on highlighting
gagement of automatic safety systems.
potential traffic conflicts ahead, and may
In such a scenario, an understanding of driver's intentions
be particularly useful for the younger inexperienced drivers,
becomes paramount. For example, the activation of the
who have been systematically found to exhibit less efficient
turn signal can temporarily disable lane departure warnvisual scanning patterns compared to experienced drivers
ings. However, cases such as intentional tailgating are dif[54], and are less proficient in interpreting the situation corficult or impossible to detect. Therefore, an open question
rectly [53], [55]. For more experienced drivers, a feedback
is: Should there be situations in which autonomous syssystem may still be helpful in directing their attention back
tems are given the authority to override driver input? It is
to critical elements in the driving scene if they are distractclear that vehicles should warn drivers of drowsiness or
ed. A number of aids have been proposed in the literature to
distraction, but it is unclear whether safety systems should
support anticipatory driving, e.g., highlighting of important
prevent intentionally risky driving behaviors. In the latcues [56], visual depiction of stopping distance [57], and
ter scenario, i.e., the transfer of control from the vehicle
visual feedback using an LED array for modulating brake
to the driver, the role of driver monitoring systems is as
pedal control [58]. However, interface designs in these studimportant but less obvious. Once the vehicle gains control
ies remain largely at the conceptual stage, and have been
(e.g., due to driver drowsiness, or at driver's request), the
evaluated in limited settings such as in driving simulators.
driver monitoring system can actively monitor the driver
by measuring facial/body expressions and physiological
Concluding remarks and future directions
signals. Such continuous monitoring of the driver allows
Recent developments in sensing technologies as well
the system to determine when it is safe for the driver to
as advancements in our understanding of human factors
regain vehicle control.
affecting driving performance provide opportunities for
A major challenge in the research and development of
the development of smart driver monitoring systems.
smart driver monitoring systems is the unique interdisciplinThese systems can use vehicle measures, facial/body exary nature of this area, which requires a close collaboration
pressions, and physiological signals to continuously monibetween researchers and practitioners in both signal processtor driver state and also sense the environment to provide
ing and human factors communities. The recent advancefeedback to drivers or take vehicle control if there is
ments in human factors research on driving safety can provide
a need. In practice, however, only vehicle measures are
insights for incorporating efficient mitigation technologies into
widely used in commercially available driver monitoring
smart monitoring systems, using both real-time and postdrive
systems. The use of facial/body expressions is limited due
feedback. This article presented a high-level interdisciplinary
to technical challenges in acquisition and real-time prooverview of the smart driver monitoring systems to facilitate
cessing of these signals. For physiological measures, the
such collaborations.
most important challenge has been nonintrusive signal
acquisition, which has not been possible for signals such
as EEG and ECG until recently. To fully exploit the inforAcknowledgment
mation provided by these three measurement categoWe acknowledge support by the Natural Science and Engiries, the information extracted from different measures
neering Research Council through the strategic partnership
can be further processed using data fusion techniques.
grant titled DREAMs: Enhancing Driver Interaction with
However, considering the fact that simultaneous real-time
Digital Media through Cognitive Monitoring.
46

IEEE SIgnal ProcESSIng MagazInE

November 2016

http://www.nhtsa.gov/staticfiles/rulemaking/

Table of Contents for the Digital Edition of Signal Processing - November 2016