Stefan Oniga

Professor

University of Debrecen
Country flagHungary

Research Interests

ApplyKite TwinLearns as you apply

Make your next move.

Prepare your application

Review drafts before you send them.

Explore related searches

Articles (17)

Year:

2026

Year:

2026

This investigation examines the Dead Hot Map (DHM) method and timing calibration for Run 14 Au+Au collisions in the PHENIX experiment. The DHM method guarantees data integrity by identifying and omitting defective detector towers (nonfunctional, hot, and very hot towers) via a set of criteria and statistical evaluations. This procedure entails hit distribution analysis, pseudorapidity adjustments, and normalization, resulting in an enhanced map of functional detector components. Timing calibration mitigates the issues associated with time-of-flight measurement inaccuracies, such as slewing effects and inter-sector timing differences. Numerous corrections are implemented, encompassing slewing, tower-specific offsets, and sector-by-sector adjustments, resulting in a final resolution of 500 picoseconds for the electromagnetic calorimeter. These calibrations improve the accuracy of photon and π0 measurements, essential for investigating quark–gluon plasma in high-energy nuclear collisions.

Year:

2025

Artificial intelligence has increasingly taken over various aspects of daily life, resulting in the proliferation of smart devices and the development of smart living and working environments. One significant domain within this technological advancement is human activity recognition, which includes a broad spectrum of applications such as patient monitoring and supervision of children’s activities. In this research, we endeavor to design a human activity recognition system that effectively analyzes multi-user data through a machine learning framework centered on graphical plot images. The proposed methodology uses a PIR sensor-based system. This system uses a two-stage process; the first one involves generating new image datasets as density map images and graphical representations based on the Kyoto CASAS multi-user dataset. In the second stage, the generated data are provided to a sequential convolutional neural network, which predicts the 16 activities developed by two users. To generate the new datasets, we only used data from ambient sensors, which were organized in windows. We tested many types of window dimensions and extra features such as temporal aspect and the limitation of two activities in one window. The neural network was optimized by increasing the deconvolutional layers and adding the AdamW optimizer. The results demonstrate the viability of this method, evidencing an accuracy rate of 83% for multi-user activity and an accuracy rate of 99% for single-user activity. This study successfully achieved its objective of identifying an efficient activity recognition methodology and a data image representation. Furthermore, future enhancements are anticipated by integrating data sourced from PIR sensors, with information gathered from user-personal devices such as smartphones. This approach is also applicable to real-time recognition systems.

Year:

2025

Year:

2025

Year:

2025

Year:

2025

In brain–computer interface (BCI) systems, motor imagery (MI) electroencephalogram (EEG) is widely used to interpret the human brain. However, MI classification is challenging due to weak signals and a lack of high-quality data. While deep learning (DL) methods have shown significant success in pattern recognition, their application to MI-based BCI systems remains limited. To address these challenges, we propose a novel deep learning algorithm that leverages EEG signal features through a two-branch parallel convolutional neural network (CNN). Our approach incorporates different input signals, such as continuous wavelet transform, short-time Fourier transform, and common spatial patterns, and employs various classifiers, including support vector machines and decision trees, to enhance system performance. We evaluate our algorithm using the BCI Competition IV dataset 2B, comparing it with other state-of-the-art methods. Our results demonstrate that the proposed method excels in classification accuracy, offering improvements for MI-based BCI systems.

Year:

2024

In this paper, we present the implementation of an artificial intelligence health assistant designed to complement a previously built eHealth data acquisition system for helping both patients and medical staff. The assistant allows users to query medical information in a smarter, more natural way, respecting patient privacy and using secure communications through a chat style interface based on the Matrix decentralized open protocol. Assistant responses are constructed locally by an interchangeable large language model (LLM) that can form rich and complete answers like most human medical staff would. Restricted access to patient information and other related resources is provided to the LLM through various methods for it to be able to respond correctly based on specific patient data. The Matrix protocol allows deployments to be run in an open federation; hence, the system can be easily scaled.

Year:

2024

This paper comprehensively reviews hardware acceleration techniques and the deployment of convolutional neural networks (CNNs) for analyzing electroencephalogram (EEG) signals across various application areas, including emotion classification, motor imagery, epilepsy detection, and sleep monitoring. Previous reviews on EEG have mainly focused on software solutions. However, these reviews often overlook key challenges associated with hardware implementation, such as scenarios that require a small size, low power, high security, and high accuracy. This paper discusses the challenges and opportunities of hardware acceleration for wearable EEG devices by focusing on these aspects. Specifically, this review classifies EEG signal features into five groups and discusses hardware implementation solutions for each category in detail, providing insights into the most suitable hardware acceleration strategies for various application scenarios. In addition, it explores the complexity of efficient CNN architectures for EEG signals, including techniques such as pruning, quantization, tensor decomposition, knowledge distillation, and neural architecture search. To the best of our knowledge, this is the first systematic review that combines CNN hardware solutions with EEG signal processing. By providing a comprehensive analysis of current challenges and a roadmap for future research, this paper provides a new perspective on the ongoing development of hardware-accelerated EEG systems.

Year:

2024

Year:

2024

Year:

2023

Year:

2022

Year:

2022

Year:

2022

Year:

2021

Year:

2020

Collaborators (1)

Balazs Ujvari

professor

University of Debrecen

HUNGARY
View details