ALGORITHMS AND SOFTWARE ARCHITECTURE OF THE INFORMATION SYSTEM FOR NEURAL NETWORK ANALYSIS OF PEOPLE POSES

Abstract

The article provides an overview of the current state of automated detection of people poses disturbances using visual data. The relevance of the study is due to the rapid increase in the number of cases of spinal curvature due to a sedentary lifestyle, prolonged work at a computer, and lack of control over body position. This necessitates the creation of innovative solutions capable of providing early diagnosis of postural disturbances using computer vision methods. The approach proposed in the study is based on the analysis of the people skeletal structure based on images with the subsequent application of mathematical and algorithmic analysis of the coordinates of key points of the spine to detect typical pose disturbances.

An applied implementation of the described mathematical and algorithmic support based on the designed information system is presented. The applied software implementation works on the basis of the YOLO neural network model, which was further trained on the specialized dataset. The model training lasted 25 epochs and allowed to achieve an accuracy of detecting key points of pose in the photo of more than 80%. The software implementation was used to validate the method, where the results of automated analysis were compared with expert assessments, which confirmed the high accuracy and sensitivity of the approach. Further research is planned to focus on the use of deep neural networks for real-time pose recognition based on a video stream. The proposed system has the prospect of widespread implementation in mobile applications and computer vision platforms for the mass user.

The importance of solving this problem is not limited to the medical aspect. Detection and correction of pose disturbances are directly related to the achievement of the Sustainable Development Goals (SDGs), in particular in the context of ensuring universal access to quality health care, improving physical well-being and introducing innovations in healthcare systems. The integration of such technologies supports both individual well-being and equitable access to healthcare services for a broad population, which is especially relevant for resource-limited countries. Thus, the developed solution has the potential to make a significant contribution to achieving global sustainable development goals, including strengthening the health of nations, developing digital medicine, and reducing inequalities in access to innovative diagnostic tools.