Работы автора

Chronic non-communicable diseases account for more than 70% of global mortality statistics. The main share is made up of diseases of the cardiovascular system. Adequate preventive measures—impact on controllable and conditionally controllable risk factors—can reduce the contribution of these diseases to the structure of mortality. A significant effect can be achieved with an adequately selected level of physical activity, but doctors do not always recommend specific actions to patients. This article describes a prototype of a cognitive assistant for constructing personalized plans for therapeutic physical exercises for relatively healthy people and people suffering from cardiovascular diseases. The developed system consists of two main components: a cardiovascular risk assessment module and an exercise planning module. The risk assessment module consists of a knowledge base and an argumentative reasoning algorithm. Its task is to identify risk factors and levels, which is dual in nature: in the case of monitoring a relatively healthy user, the risk of developing cardiovascular disease is assessed, while in the case of interaction of the system with a user with cardiovascular disease, the risk of complications of a chronic form is assessed—development of a cardiovascular event. The exercise planning module includes an exercise database and a scheduler algorithm. The planning algorithm selects optimal therapeutic physical exercises according to optimal criteria, in order to form a plan that will not harm the patient and will increase his physical performance. The developed mechanism allows you to create training scenarios for users with any level of initial training, taking into account the available sports equipment, the preferred location for training (home, street, gym) and at any level of the cardiovascular continuum.

The paper presents a new multimodal approach to analyzing the psycho-emotional state of a person using nonlinear classifiers. The main modalities are the subject’s speech data and video data of facial expressions. Speech is digitized and transcribed using the Scribe library, and then mood cues are extracted using the Titanis sentiment analyzer from the FRC CSC RAS. For visual analysis, two different approaches were implemented: a pre-trained ResNet model for direct sentiment classification from facial expressions, and a deep learning model that integrates ResNet with a graph-based deep neural network for facial recognition. Both approaches have faced challenges related to environmental factors affecting the stability of results. The second approach demonstrated greater flexibility with adjustable classification vocabularies, which facilitated post-deployment calibration. Integration of text and visual data has significantly improved the accuracy and reliability of the analysis of a person’s psycho-emotional state

В статье рассматривается проблема прогнозирования деятельности агента исходя из текстового описания задачи и визуального анализа среды. Предложено обновление подходов классической когнитивной архитектуры, позволяющее применять её в реальной среде. Разработано дополнение семиотического метода символьного обозначения авторским нейросетевым механизмом связывания векторов текстового и визуального пространств. Проведен ряд экспериментов с полученной моделью в комплексной среде эмулятора вождения автомобиля.