Gaze-Controlled Turkish Virtual Keyboard Application with Webcam

Abstract

In this study, an easy-to-use and inexpensive system aimed at solving the communication needs of people with disabilities such as ALS by writing with their eyes was introduced. The performance of the system are measured by writing speed (word per minutes) and error rate. The system consists of a webcam that fixed in front of the eye with a plastic eyeglass, and a computer. On the keyboard proposed in this study, the characters to be typed are placed at 22 discrete points which can be detected without error. 16 characters which are most used in Turkish (total %85 usage rate), can be selected with a single gaze. Other characters can be selected with gazing these points with a dwell time. The iris circle on the image recorded by the camera is found according to the Hough transform and the position of the center coordinates of the eye is classified by using the k-nearest neighbor algorithm from the machine learning methods and the character to be written is determined. In order to select the letters at the gaze points, the letter that is highlighted by the software in voice and color is moved to the writing area in the middle of the screen by eye. As a result of the first experiments, it has been observed that the system developed allows the average of 6.63 words per minute (approximately 33 letters) to be written with an error rate of 0.86%.