Modeling and Research of User Gaze Detection Methods in Human-Computer Interaction Systems

The article considers approaches to organizing contactless human-computer interaction based on gaze direction. The architecture of the system is presented, consisting of modules for face detection, head orientation determination, eye position analysis, and gaze direction prediction using a neural network model. The importance of the stage of pre-processing and normalization of images is emphasized to ensure stable operation of the system under changing ambient lighting and user head positions. The use of a neural network that receives normalized eye and face images as input and generates a prediction of gaze direction in the form of tilt angles is highlighted. It is noted that this data is converted into a three-dimensional vector that intersects the screen plane to determine the coordinates of the point at which the gaze is directed. The developed mechanism for confirming a choice using a double blink is also described. This solution allows you to execute commands without the need to use a keyboard or mouse, which is especially relevant for users with physical disabilities. As a result of the study, it was concluded that the developed system demonstrates high accuracy and stability when working in real time. The proposed solutions open up prospects for the further development of systems for natural human-computer interaction, in particular in the field of assistive technologies.