1. Drowsy Driving Detection
and Alert System
Patrick Coughlin
David Rubenstein
Karan Singh
Department of Electrical and Computer Engineering
2. Motivation
• Each year, the National Highway Traffic Safety Administration
(NHTSA) estimates that 100,000 police-reported accidents
are the result of drowsiness
• This includes at least 1,550 deaths and 71,000 injuries, as
well as $12.5 billion in total monetary loss
• 60% of adult drivers have said
that they have driven while
drowsy, and 37% have said that
they have fallen asleep behind
the wheel
3. Introduction
• The purpose of the Drowsy Driving Detection and Alert
System is to aim to reduce the number of drowsiness-related
car accidents
• The system accepts biological and behavioral signals from the
driver, and alerts the driver if it determines that he or she is
drowsy
4. Design Overview
• The system is composed of four cooperative subsystems:
• Steering wheel attachment
• Detects grip on steering wheel and the return-to-center steering
tendencies of the driver
• Seat belt attachment
• Detects declining breath rate of the driver
• Dashboard camera
• Captures images of the driver’s face in order to determine eye
closure
• Control unit
• Receives data from sensor systems and continuously executes
drowsiness-detecting algorithm
5. Drowsiness Conditions
• Gyroscope System
• Considered drowsy when a 3% increase between samples
is detected
• Force System
• Considered drowsy when a 20% increase between samples
is detected
• Breath Rate System
• Considered drowsy when breath rate falls below 12
breaths per minute
• Camera System
• Considered drowsy when eyes are at least 80% closed for
10% of a minute (state of perclos)
6. System Prototype
• Blue box contains power source, speaker, and all other circuit
components
(not pictured: breath rate stretch sensor wraps tightly around driver’s waist)
7. Test Results
• Subsystem Testing
• Each individual subsystem was tested to ensure that the drowsiness
algorithms were properly integrated and calibrated for sensitivity
issues
• Full System Testing
• System can correctly alert a driver in proper lighting
• Additional testing includes system in darkness and a driver wearing
glasses, which both failed
8. Future Work
• Similar products are currently being developed at major car
manufacturers, such as Volvo and Mercedes Benz
• These products monitor the same drowsiness symptoms as our
Drowsy Driving Detection and Alert System
• This system could either be used as a product that is
integrated within a car’s architecture, or as a research tool to
help determine conditions under which drowsiness occurs
while driving, since it can be easily recalibrated
• There is a definite and immediate need for this type of
product
Steering Wheel subsystem consists of a Gyroscopic analog velocity sensor and 2 force sensitive resistors that are long strips on either side of the steering wheel. These sensors are used to detect a driver’s grip strength and steering corrections. A sudden change in grip and a sudden steering correction can indicate that a driver is drowsy.
There is a breath rate sensor made of a conductive rubber cord that is wrapped around the user’s body. A decreased breath rate also indicates drowsiness.
A small rig that you can see here was made to simulate driving. There is a box that contains all of the circuit components, the raspberry pi, and a speaker. The speaker plays a tone 3 times when the driver is determined to be drowsy to hopefully wake them up.
Ideally this system would simply be a program that is run on the car’s main computer and the car’s audio system would be used to alert the driver when they are drowsy. The installation would be a part of the manufacturing of the car and drowsiness thresholds would be a part of the car’s overall software design.
The system could also be used as a research tool. If a car was retrofitted to include the sensors we chose the system could be used during actual driving. The drowsiness detection thresholds would likely need to be changed based on the individual car and this would require a custom installation.
Now we would like to move onto the demo.