Today, an urgent scientific task is to develop an information technology for identifying changes in the emotional state of a person by facial expressions, which will increase the accuracy of identifying abrupt changes in the emotional state in a video stream in real time, which will thus improve the process of detecting abnormal behavior of a group of people in a crowd for systems that meet security requirements. As a result of the dissertation, the actual scientific and applied problem of improving the process of detecting abnormal behavior of a group of people in a crowd by their facial expressions in systems that meet security requirements was solved.
The object of research is the process of detecting abnormal behavior of a group of people in a crowd by their facial expressions in systems that meet safety requirements.
The subject of the research is models, methods and means of information technology for identifying changes in the emotional state of a person by facial expressions for systems that meet security requirements.
The purpose of the dissertation is to improve the accuracy of identifying changes in the emotional state of a person by facial expressions by developing information technology to detect abnormal behavior of a group of people in a crowd by their facial expressions in systems that meet security requirements.
The scientific novelty of the obtained results is as follows:
1) a new model of representation of facial expressions of human emotional states was developed, which, unlike analogues, stably groups and separates the main classes of emotions, which made it possible to use low-resolution images in video surveillance cameras and detect sudden changes in emotional state;
2) a new method of geometric interpretation of facial areas was developed, which, unlike analogues, makes it possible to transparently obtain characteristic features of facial activity, which made it possible to analyze low-resolution images with low computational complexity;
3) the method of hyperplane classification for identifying facial expressions of emotional states was improved, which, unlike analogues, allows to build a hyperplane of separation in the vector space of features on the principle of «man in a loop», which made it possible to obtain classifiers for detecting sudden changes in emotional states;
4) the information technology for identifying abrupt changes in emotional states was further developed, which allowed for localizing groups of people with abrupt changes in emotional states based on external video recording materials with a high level of accuracy.
The combination of a new model of representation of facial expressions, a new method of geometric interpretation and an improved method of hyperplane classification in information technology made it possible to obtain a high accuracy of classification of human emotional states (up to 82.42%), which provides security personnel with a reliable and effective tool for understanding crowd dynamics and predicting potential security risks during mass gatherings. The results of experimental testing using the developed software prototype confirm the validity of the scientific provisions of the proposed information technology, since its implementation makes it possible to increase the reliability of detecting abnormal behavior by facial expressions by 0.91-2.20%, depending on different emotions and environmental conditions, and to reduce the likelihood of errors in identifying sudden changes in emotional states by 0.23%-2.21% compared to modern analogues.
Theoretical and practical results of the research are implemented in PE “Shelter Plus” (Khmelnytskyi), LLC “ITSYTS” (Khmelnytskyi) and in the educational process of Khmelnytskyi National University during the teaching of disciplines at the Department of Computer Science for the specialty 122 Computer Science, as well as in the implementation of research work on two state budget topics of Khmelnytskyi National University “Agent-based system for improving the security and quality of computer system software” and “Development of information technology for making human-controlled critical and security decisions based on mental and formal models of machine learning”, in which the author of the dissertation was a direct performer.
