Signal Processing - November 2016 - 38

Smart Driver Monitoring
Physiological Measures

Vehicle
Measures

Skin Conductivity
Respiratory Activity
Heart Activity
Brain Activity

50
50

Driver Feedback

Steering Angle
Throttle Position
Brake Position
Braking Intensity
Vehicle Heading
Acceleration
Velocity
Headway Distance
Speed Limit Detection
Over/Under Speeding
TLC
Lane Detection
LD
Jerk

CPU

Audio
Visual
Vibration
Break/Rest Reminder
Device Lockout
Adaptive Cruise Control
Collision Avoidance
Lane Keeping Assistance
Facial and Body Expression
Gaze Direction
Eye Movements
Eye Blinks
Eye Tightening/Closure
Yawning
Head Movements

FIGURE 2. An illustration of sensors and feedback technologies that can be used in smart driver monitoring systems. TLC: time-to-lane crossing.

number of vehicle measures are affected. Over/underspeeding
happens more frequently at higher levels of sleepiness, and so
does large LD variance. Moreover, during interval G and at
the end of interval I, the driver changes lanes very frequently,
which suggests a degradation in lane keeping ability for this
particular driving scenario. Finally, at higher levels of sleepiness, TLC tends to get closer to zero, which is an indicator
of unsafe driving. This effect is reflected in the graphs that
indicate the number/fraction of TLC measure going below a
certain threshold. Note that the experimental setup used for
the study in Figure 3 is described in Figure 4. A comprehensive review of the literature on vehicle measures is provided
in [13] and [19].
Almost all major automobile manufacturers have
equipped their newer high-end models with warning technologies, such as lane departure (which tracks the vehicle's
position in the lane and alerts the driver when the vehicle
starts to drift laterally) and collision warning (which uses a
forward-looking camera or a radar to measure the distance
to the lead vehicle and alerts the driver when this distance
is less than a safe threshold). Furthermore, many manufacturers such as Tesla, Mercedes-Benz, BMW, Lexus,
Infinity, and Honda have equipped some of their models with vehicle control intervention technologies. Some
examples include lane-keeping assistance (which actively
applies corrective torque to the steering wheel to keep the
vehicle in the lane) and collision avoidance (which activates the brakes if the driver does not respond to the collision warning).
Aside from the aforementioned systems, some manufacturers have introduced smart technologies that utilize vehicle
38

measures to monitor driver states, such as drowsiness. For
example, Mercedes-Benz's Attention Assist (http://m.mercedesbenz.ca/en_CA/attention_assist/detail.html) warns the driver
if signs of drowsiness are detected. This system is mainly
based on tracking steering wheel movements, while assessing
parameters such as speed, longitudinal/lateral acceleration,
and indicator and pedal usage. It also takes into account external factors such as crosswinds and the unevenness of the road.
Another example is Volkswagen's Driver Fatigue Detection
System (http://www.volkswagen.com.au/en/technology_and_
service/technical-glossary/fatigue-detection.html), which continually measures steering wheel movements together with
other vehicle signals as side information, and warns the driver
if signs of fatigue are detected.

Facial and body expression measures
Observable cues from driver's facial and body expressions can
be used to infer about a driver's arousal level. For example,
eye blinks, percentage of eye closure (PERCLOS) [20], and
yawning are particularly useful for detecting fatigue and
drowsiness, whereas eye movements, gaze direction, and head
movements provide information on drivers' visual attention
(e.g., on the road or infotainment system). These measures are
heavily reliant on video processing, and thus have been traditionally difficult to implement due to computational limitations. However, recent advancements in computational power
and algorithm design have made such implementations possible in real time.
Figure 4 illustrates an example of using video cameras
[in (a), a dashboard-mounted camera is shown and in (b) the
head-mounted eye tracker is shown] in a research setting

IEEE SIgnal ProcESSIng MagazInE

November 2016

http://www.m.mercedes http://www.benz.ca/en_CA/attention_assist/detail.html http://www.volkswagen.com.au/en/technology_and_

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