University of Waterloo researchers have devised computer algorithms that can accurately determine when drivers are texting or engaged in other distracting activities. This is based on optics, with cameras detecting hand movements that are regarded as deviations from normal driving behavior. The signals are fed into an artificial intelligence system that classifies the hand movements and responds if the hand movements are deemed to be potential safety threats.
The platform was devised by a team led by Professor Fakhri Karray. The aim of the device is to improve road safety through issuing alerts to drivers when they appear to be distracted. The aim is to add this to vehicles to be used by businesses and consumers.
The technology would also fit in with self-driving cars since a signal that a driver has become dangerously distracted would allow the car to take over driving under the most serious circumstances (such as a risk of crashing).
For business vehicles travelling in convoy, this system could connect with the developing Internet of Cars.
The algorithms that form the basis of the technology were trained by machine-learning techniques. These techniques ‘learn’ driver activities like texting or making calls not made using hands free functions. Other recognized activities of concern include fidgeting or reaching for objects.
In addition to movement, the technology also accounts for the duration of the activity. The algorithms are built upon research into computer recognition of signs of drivers seemingly falling asleep at the wheel. For this, head and face positioning are also important cues.
Longer term, the researchers hope to combine the detection software for drivers falling asleep with the platform that notes texting, into one system covering all aspects of driver distraction. This may extend to facial recognition analysis, based on eye blink rate, to determine whether a driver is paying sufficient attention while driving.
Commenting on this, Professor Karray notes how distracted drivers are a risk and the activities have “a huge impact on society.” The researcher quotes figures that indicate distracted drivers are to blame for some 75 percent of all global traffic accidents.
The research findings were recently presented at the 14th International Conference on Image Analysis and Recognition in Montreal, which took place during July 2017. The conference focused on the fields of image processing; image analysis; and pattern recognition.
