Object: processes of human-machine interaction that identify, track and classify hand gestures or other objects in the image and video sequence; objective: improving the characteristics of automatic recognition of static and dynamic hand gestures due to the development and practical use of models and methods of information technology of human-machine interaction with the use of gestures is set and solved in the work; methods: set theory, graph theory, matrix theory, submodularity principle, greedy heuristics, annealing simulation method, entropy maximization, Voronoi diagram, Delaunay triangulation - for placement of sensors and IoT devices in the monitoring system; probability theory, methods of descriptive statistics, variational calculations, correlation analysis - for the methodology of long-term data processing and analysis of water quality; principal components method, factor analysis - to determine a set of sensors in IoT water quality control devices; SCAI-graph, mashup methodology, prototyping technologies, compression measurement models - when creating information technology for designing surface water monitoring systems based on IoT; novelty: for the first time, a methodology for developing and sharing visual hand gesture recognition techniques has been developed to create and explore deep learning models for recognizing static and dynamic gestures capable of real-time operation and to understand how to configure them for different gesture control interfaces and potential applications in HMI systems; the model of static gesture recognition, built according to the proposed methodology based on a convolutional neural network, by artificially enlarging data and using contours, has been improved. Thanks to the use of contours, the model is resistant to relatively wide angles of rotation of the hands and independent of lighting; gesture recognition and prediction technology based on the sequence generation model using ConvLSTM2D and Conv3D has been further developed; the model of the finite state machine for contactless control of viewing of medical images by means of gestures which, unlike existing, uses the data of the predicted frames of video sequences that allows to reduce response time of system has been improved; a new dataset has been created to test the proposed models and method of information technology for solving problems of gesture recognition and prediction in the operating room, and conceptual development of an intuitive vocabulary of dynamic gestures was created, which allows to implement an effective contactless interactive system adapted to the surgical context; the structural model of HMI information technology using gestures has been improved by identifying the main stages and information flows of creating and integrating deep learning models, which provides decisions on the application of developed methods, tools and technologies; research results: software implementations of models of the recognition system have been developed, which demonstrates high accuracy and speed with low sensitivity to lighting conditions. The results of the dissertation are used as the basic elements of the HMI system for navigation and viewing medical images in the operating room.
