Abstract
We developed a method for predicting drowsiness in drivers based on changes
in the heart rate. Previous research focused on detecting the napping propensity of drivers.
However, the detection time for that method is often too long to prevent napping because
drivers may already feel drowsy by the time detection is complete. The ability to predict
drowsiness early when drivers do not yet feel drowsy, even if they show signs of drowsiness,
is useful for preventing drivers from falling asleep at the wheel.
The heart rate is affected by the ambient temperature, the posture of drivers, and their mental
state. Drowsiness is the critical factor for drivers because most of the time ambient
temperature or posture do not change much while driving. Furthermore, a sign of drowsiness
is a decreasing heart rate.
To measure the heart rate of drivers easily, we placed electrodes on both sides of a steering
wheel and devised a method for predicting drowsiness based on changes in the heart rate.
These electrodes were made by painting electro-conductive adhesive, which includes silver,
on a steering wheel or a steering wheel cover. The heartbeat is measured by amplifying the
electronic potential difference between both hands. However, the noise level increased when a
heartbeat was detected through these electrodes, as compared to using regular electrodes. This
is because of the unstable contact between hands and electrodes due to the behaviour of hands
or vehicle vibrations. So, we improved the detection performance by computing the mutual
correlation between the heartbeat signal and a reference peak pattern to reduce the influence
of noise. In cases when the heartbeat was impossible to detect due to steering using only one
hand, we interpolated a predicted value that was computed from previous data. By using this
interpolation, predicting drowsiness is possible while the ratio of the amount of time spent
steering with both hands to the total steering time is over 50%..
To evaluate the reliability of our method, we conducted an experiment on a highway with ten
subjects. Overall, our method was 70.7% reliable for predicting drowsiness. We will improve
the reliability by optimizing the sensitivity of the sensor for different individuals.
Keywords- Safety, Preventive Safety, Drowsiness/Sleepiness, Driver, Heart rate