Currently, it seems that the trend that is set by the main mobile manufactures is to narrow the physical man-machine interface
Abstract
Currently, it seems that the trend that is set by the main mobile manufactures is to narrow the physical man-machine interface, and progress to other control methods, such as voice control and vision control.
There is still no complete solution for controlling devices with just the eyes.
This project is based on a paper published in 2009 at the “Machine Vision and Applications” journal titled “Single image face orientation and gaze detection”. This work was done by Prof. Jeremy Yrmeyahu Kaminski, Dotan Knaan, and Adi Shavit. This paper was chosen in order to check the feasibility of software implementation for detection of human gaze.
In this project, we have implemented face features detection and a gaze calculation mechanism.
The results are inconclusive, but point out on a number of obstacles in turning this to a ready-to-sell product: eye feature detection (especially in dark corneas), unnatural environment (LED is needed in order to create a glint), low resolution of front cameras, structural and physiological differences between people and more.
The algorithm still needs to be improved in order to become more robust, but it shows good enough results in some cases and for some uses.
Flowchart
Results
