Gesture recognition using Kinect to detect unsafe behaviors performed by drivers

Abstract

Vehicle drivers’ unsafe behaviors, like using a cellular phone while driving, is one of the leading
causes of traffic accidents worldwide. This unsafe practice is hard to regulate and control.
However, it is possible with the use of an automated gesture detection system that indicates a
dangerous behavior. The use of cameras with 3D depth sensors can aid in analyzing objects
and people in real time with high precision, identifying features and figures. These sensors use
a technology referred to as “point cloud”, which emits an array of laser lights and measures the
time of return (fly time) back to the sensor, determining the depth of each one of these points.
This technology permitted the development of a gesture recognition application that is capable
of automatically detecting the action that indicates that a driver is using a cellular phone while
driving. An initial functional simulation was developed using background subtraction, feature
analysis, and histogram analysis algorithms using the program V-Rep, which can detect the
gesture that would indicate this behavior. Furthermore, a prototype was implemented using real
images taken with the Kinect that were processed using Octave using the same algorithms used
for the simulation, thus supporting the effectiveness of the application.