Search academic texts

Information × Registration Number 0225U002815, (0124U004650) , R & D reports Title Non-Gaussian probabilistic models for face recognition. popup.stage_title Побудова негаусiвських ймoвipнicних моделей для розпiзнавання облич. Head Prykhodko Serhii B., Доктор технічних наук Registration Date 07-04-2025 Organization Admiral Makarov National University of Shipbuilding popup.description1 Increasing the probability of face recognition due to the use of non-Gaussian probabili models. popup.description2 Research Object – Face recognition. Research Objective – Improving the reliability of face recognition through the application of non-Gaussian probabilistic models. Research Methods – Methods of probability theory, mathematical statistics, mathematical modeling, and multivariate statistical analysis of systems with non-Gaussian random variables. Scientific Novelty of the Obtained Results: For the first time, a probabilistic model for face recognition has been developed in the form of a prediction ellipsoid with an F-distribution quantile for the normalized feature vector. Unlike existing models, it accounts for deviations of the data distribution from the Gaussian distribution by applying a ten-dimensional Box-Cox transformation and considers the number of data points through the use of the F-distribution quantile, which enhances the reliability of face recognition. An improved probabilistic model for face recognition has been developed in the form of a prediction ellipsoid with a Chi-square quantile for the normalized feature vector. Unlike existing models, it accounts for deviations of the data distribution from the Gaussian distribution by applying a ten-dimensional Box-Cox transformation, which improves the reliability of face recognition. Forecast Assumptions Regarding the Development of the Research Object – The development of information technology and an automated system for constructing non-Gaussian probabilistic models. Such non-Gaussian probabilistic models can also be used for recognizing other patterns, such as keystroke dynamics.   Product Description popup.authors Prykhodko Serhii B. Trukhov Artem S. popup.nrat_date 2025-04-07 Close