The problem of gaze detection has received a huge amount of attention in the past decade. Wide variety of commercial products are already available in the market
The problem of gaze detection has received a huge amount of attention in the past decade. Wide variety of commercial products are already available in the market, whether in “human like” robot interaction systems, car driver safety applications or even smart phone applications. Consequently, there are many proposed algorithms for gaze detection. Most conventional methods require use of special hardware such as high resolution cameras or IR-LEDs to detect key features in the face and perform complex geometrical calculations to determine the gaze direction.
This project wishes to achieve a Gaze Detector which is simple to implement on any system with no need of manual user calibrations, require no more than a single web camera, is robust enough to operate in different environments and fast enough to work in real time.
In order to achieve that, the proposed Gaze Detector is based on machine learning capabilities and incorporates several supervised learning algorithms, used for training dedicated classifiers to detect the head, the eye region and determine the eyes orientation. Those classifiers must be robust and therefore trained in different environments. In addition the classifiers are combined and cascaded to enhance the overall detection performance.