Signal Processing - November 2016 - 36

driving (e.g., alerts of incidents ahead), and foster positive
in the United States found the driver as the critical reason
behavioral changes in the long run (e.g., by providing aggrein 94% of crashes, citing errors such as distraction, sleep,
gated information after a trip to help drivers understand potenexcessive speed, and false assumption of others' actions [3].
tial consequences of unsafe driving behaviors). A smart driver
Certain driver populations are at an elevated risk. Older
monitoring system can also address some of the risk factors
drivers are affected by age-related declines in perceptual,
associated with certain driver characteristics. For example, the
cognitive, and motor abilities [4], while younger and par2016 Chevy Malibu has a feature tailored to teen drivers. This
ticularly novice drivers tend to lack sufficient skills to recfeature warns teen drivers when they exceed a predetermined
ognize or anticipate road hazards [5]. Younger drivers may
speed limit, blocks the sound from
also be risk unaware and engage in
the stereo until the front seat belts are
potentially risky activities, such as
The human factors community
buckled, and generates a report card
speeding. For example, among U.S.
has made significant progress in
on driving safety metrics (e.g., numdrivers involved in fatal crashes in
understanding driver behaviors,
ber of over-the-speed-limit warnings
2013, males aged 15-24 years old
and assessed the efficacy of
and forward collision alerts). In addiwere the most likely to be speeding at
various interventions for unsafe
tion to real-time feedback on speed
the time of the crash [2]. In addition
and seat belt use, the system provides
to the driver's age and gender, there
driving practices.
an opportunity for parents to discuss
are social-psychological factors that
safer driving behaviors with teen drivers in a postdrive feedcan also contribute to crash risk. For example, self-reported
back form, thus calibrating or reinforcing a more accurate
sensation-seeking tendency, which is a personality trait, has
mental model of safe driving.
been associated with risky driving (e.g., records of impaired
While the Malibu's teen driver monitoring system provides
driving and self-reported speeding [6]). Attitudes (evaluations
useful feedback, it does not directly address the more tranof the potential outcomes) and perceived social norms (how
sient characteristics of the driver. The driver's state, such as
other people behave, or what societal expectations there are)
fatigue, anger, or mental overload (i.e., when the driver's cogabout unsafe driving behaviors can also play a role in vehicunitive capacity limit has been reached), can influence driving
lar crashes. For example, [7] found that drivers who drove
behavior, both in terms of information processing capabilifaster than others also had a more positive attitude toward
ties (e.g., reduced reaction times to hazards due to drowsiness)
speeding. The same study also found that drivers who perand decisions to perform risky behaviors (e.g., following a
ceived others to drive at excessive speeds were more likely to
vehicle too closely while in a state of aggravation). The ability
drive fast compared to those who perceived others to comply
to monitor driver state can ensure that factors such as current
with the limits. Medical conditions, such as sleep apnea [8],
mood, mental workload, and vigilance level of the driver are
are also known to increase crash risk. Finally, more transient
taken into account and feedback is provided in a timely mandriver characteristics, i.e., driver state, such as fatigue, drowsner (e.g., based on early detection of drowsiness rather than
iness, distraction, mental workload, and mood, can also affect
detection of unsafe driving performance). However, there are
safety by impairing a driver's information-processing abilimany challenges associated with monitoring driver state. The
ties and altering his or her risk-taking tendencies.
remainder of this article focuses on addressing these chalInfotainment systems and smart personal devices (e.g.,
lenges, as well as reviewing important design considerations
smartphones) can add to the already demanding activity of
regarding driver feedback. Overall, we aim to 1) provide readdriving and may increase crash risk. Digital interactions have
ers with an overview of current and emerging signal acquibecome an integral part of many drivers' daily life such that
sition and processing techniques for driver state monitoring
they expect digital content to be available even while driving.
systems, and 2) draw insights from human factors research on
Furthermore, it is inevitable that a growing array of technolodriver behaviors and intervention strategies to facilitate disgies find their way into the vehicle given their large economic
cussions about existing techniques for smart driver monitorvalue. It should be noted that although certain types of teching systems.
nology lead to distraction, technology can also enhance safety.
Automobile manufacturers are constantly seeking innovative solutions that leverage emerging sensor technology and
Driver state monitoring
increased computational power to support the driver. In the
The state of a driver can have a significant impact on her drivpast decade, there has been a rapidly growing interest in both
ing performance, a relationship best captured by the wellthe signal processing and the human factors communities to
known Yerkes-Dodson law [9]. The Yerkes-Dodson law,
develop smart driver monitoring systems that can sense and
when adapted to driver state [10], suggests an inverted-U funcmonitor the driver's state, vehicle operation, and changes in the
tion between driver's arousal and driving performance. As
environment to provide drivers with useful and appropriately
illustrated in Figure 1, both low arousal and high arousal are
timed feedback.
associated with performance decrements, and optimal perforA smart driver monitoring system can provide immediate
mance occurs when there is an appropriate amount of arousal
driving assistance (e.g., automated lane-departure warnings),
to keep the driver attentive, but not stressed. On one hand, a
support drivers in maintaining situation awareness (SA) while
driver who is fatigued or has lost vigilance due to monotonous
36

IEEE SIgnal ProcESSIng MagazInE

November 2016

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