Drivers talking or texting on their smartphones have given mobile devices a bad reputation as road distractions. But a new smartphone app could actually warn drivers if they appear distracted or drowsy behind the wheel.
The app uses smartphone technology to mimic fancier car safety systems by detecting signs of trouble that could lead to accidents. Modern phones with cameras facing both front and back allow the app to monitor a driver's head pose, eyes and blinking rate to detect possible distraction or drowsiness, even as it keeps watch on the road ahead.
"We can determine the distance between cars in front and whether a driver is changing lanes on the outside, while detecting drowsiness and distraction inside," said Andrew Campbell, a computer scientist at Dartmouth College.
The smartphone sits mounted on a dashboard holder as a hands-free helper for drivers. Whenever the car safety app detects any dangerous patterns in driver behavior or outside conditions, it warns the driver with a blinking light and a noise alert.
Such an app could provide a cheap, effective safety solution that works with any car model. But the app developers — a team from Dartmouth College, Autonomous National University of Mexico, Microsoft Research Asia and the University of Bologna in Italy — faced huge challenges while making their idea into a reality.
First, today's smartphones don't have the capability to process video streams from both the front and back cameras simultaneously. Campbell and his colleagues had to develop intelligent algorithms that switch quickly between the two cameras while processing the data. [ Best Smartphones Reviews & Comparisons ]
The rapid-switching solution means that the car safety app technically has a blind spot in the front or back at any given time. Researchers fixed that problem by using the smartphone's other sensors, such as accelerometers and gyroscopes, to figure out what is going on in the blind spot either inside or outside the car at any given time.
"When the phone camera is looking at the driver, we can use accelerometer sensors to detect if the car is weaving," Campbell told TechNewsDaily. "We use sensors to fill in when the camera is looking at the wrong place."
Second, the app's vision algorithms can push the computing limits of existing smartphones and slow down the image-processing from the cameras. If a smartphone camera typically captures video at 30 frames per second, the app's software analysis slows the process to about 8 frames per second on an Android smartphone that has a single computer processing core.
That limitation forced the developers to design vision algorithms with minimal computing power requirements as they ran tests on Samsung Galaxy smartphones. To get the best results, they plan to first market their app for the latest quad-core smartphones running Google's Android system sometime in 2013 (sorry iPhone users).
For now, the team aims to complete road testing of the application with a small pool of users in either November or December. They have enlisted 20 people and three different cars models, but admit that it's difficult to push the limits of the app without having drivers almost falling asleep behind the wheel.
"Unless someone is up all night, it's difficult to predict whether they'll be tired or not," Campbell said.