Abstract:

The transition to remote education because of the covid19 pandemic has brought many problems to educational institutions. However, in many cases, the new learning format, with a lector broadcasting the same material to all students, gives us a unique chance to collect and evaluate a variety of material parameters. Previously, during classes in the same classroom, all students were in different conditions - they looked at the lecture material from different angles, heard the material in different ways (the last rows hear the lecturer less than the first ones) and as a result, the system of lecture quality estimation using artificial intelligence and cameras in the classroom became very complex and had to take into account a huge number of parameters.

In a situation when all students are in almost identical conditions, the task (and as a consequence, the accuracy of the system) is greatly enhanced. However, these are not all the advantages! It is possible to introduce the possibility of estimating the material by the students right during the lecture. All this gives us many new parameters for material quality estimation.

In this article, the authors want to present a project of a system that will monitor the student's picture in close to real-time using a video camera, and based on the coordinates obtained from the system to determine the direction of the gaze, to identify the saccades. Saccades facilitate the creation of schemes of interaction between the student and the information source by selecting sets of saccades according to special mathematical metrics. In turn, the interaction schemes are a valuable source of useful data describing the order, speed, and integrity of the process of learning certain material. Such parameters concerning a single piece of material will make it possible to determine its average complexity, its relative and average familiarity, the size occupied concerning lecture time, and so on. These metrics will be indispensable in the process of creating and adjusting the curriculum.

The concept of this project is based on artificial intelligence and big data technologies. The recognition and classification of images will allow us to highlight the necessary data regarding the student's gaze. Manually composed data sets of large amounts of text and simultaneous recording of a student's image while reading it will provide us with data for training neural networks.

An important aspect that affects the entire system is the fact that all students have different Internet connection speeds and channel quality. This can cause the system to fail or be very slow and inaccurate. To reduce this negative impact, the authors of the article suggest several ways to significantly reduce the amount of transmitted information without affecting the quality of the system itself. It should also be noted that the presented system is based on the subsystem of gaze direction recognition, which was developed by the authors of this article.

Acknowledgement:
Part of discussed in this paper results were obtained in the framework of the Research Project titled "Component's digital transformation methods' fundamental research for micro- and nanosystems" (Project #0750-2020-0041) financially supported by the Russian Federation Ministry of Science and Higher Education.

Keywords:

Artificial intelligence, material quality, mathematical metrics, quality estimation.