Infrastructure for the Real-Time Biosignal Datasets Collection
Вантажиться...
Дата
2026
Науковий керівник
Назва журналу
Номер ISSN
Назва тому
Видавець
КПІ ім. Ігоря Сікорського
Анотація
The effectiveness of modern methods of biosignal analysis largely depends on the availability of structured data sets containing both primary signals and related metadata. At the same time, most existing biosignal registration systems do not provide automated accumulation of such data in centralized databases, which complicates the formation of training samples and the further application of machine learning algorithms. The article considers an automated system for forming representative sets of biomedical data to increase the efficiency of machine learning tasks in the analysis of the human physiological state. The infrastructure of the information system for automated collection of biosignals and metadata and formation of data sets in real time has been developed. The system has a multi-level architecture that includes a sensor level for registering biosignals, a communication level for data transmission, and a server level for their processing and storage. Signals were registered during system testing using a photoplethysmographic sensor integrated with a wireless module based on a microcontroller with Wi-Fi support. The server part is implemented in a virtualized environment using open source software, a web server, a database management system, and signal processing software modules. For visualization of biosignals and interaction with users, a Web-API and a web interface have been developed, which provide access to measurements, metadata management, and signal visualization. The system implements a data streaming pipeline that includes query verification, signal storage, and biomedical parameter calculation to assess the user’s functional state. An experimental study of the system’s performance was conducted in a load testing mode that simulates the simultaneous operation of a significant number of sensor devices. The results showed that the developed infrastructure is capable of handling more than a hundred simultaneous connections with an average query processing time of less than 100 ms. The results obtained confirm the possibility of using the proposed system for scalable collection of biosignals and the formation of data sets suitable for further application of machine learning methods.
Опис
Ключові слова
data collection automation, data set generation, Internet of Things, biosignal registration, data streaming processing, functional state determination, автоматизацiя збору даних, Iнтернет речей, реєстрацiя бiосигналiв, потокова обробка даних, визначення функцiонального стану, генерацiя наборiв даних
Бібліографічний опис
Mosiichuk, V. S. Infrastructure for the Real-Time Biosignal Datasets Collection / Mosiichuk V. S., Sharpan O.B., Yalosovetskyi V. I. // Вісник НТУУ «КПІ». Радіотехніка, радіоапаратобудування : збірник наукових праць. – 2026. – Вип. 103. – С. 78-84. – Бібліогр.: 18 назв.