Information und Kommunikation
Refine
Year of publication
Document Type
- conference proceeding (article) (183)
- Article (79)
- Part of a Book (12)
- conference proceeding (presentation, abstract) (11)
- Preprint (7)
- conference talk (4)
- Book (3)
- Report (3)
- Bachelor-/Diplom Thesis (2)
- Working Paper (2)
Is part of the Bibliography
- no (310)
Keywords
- speech synthesis (5)
- Chief Information Officer (4)
- Cloud Computing (4)
- Software (4)
- incremental processing (4)
- prosody (4)
- spoken dialogue systems (4)
- Insourcing (3)
- Literaturbericht (3)
- Magneto-optics (3)
Institute
- Fakultät Informatik und Mathematik (213)
- Fakultät Elektro- und Informationstechnik (89)
- Laboratory for Safe and Secure Systems (LAS3) (31)
- Labor Regensburg Strategic IT Management (ReSITM) (23)
- Labor Intelligente Materialien und Strukturen (8)
- Fakultät Angewandte Natur- und Kulturwissenschaften (6)
- Labor Elektroakustik (6)
- Labor Hochfrequenztechnik (6)
- Labor Optische Übertragungssysteme (6)
- Labor Informationssicherheit und Complience (ISC) (5)
Begutachtungsstatus
- peer-reviewed (150)
- begutachtet (8)
As autonomous driving becomes increasingly feasible, the German government has introduced a legal framework to enable the operation with Level 4 automated driving functionality. A key requirement is the maintenance of a continuous connection between such vehicles and a remote technical supervisor. If this link is lost, the vehicle must transition into a safe state by bringing itself to a controlled stop. To mitigate the risk of connection loss, accurate forecasting of mobile network availability along routes is essential. This paper presents an Exploratory Data Analysis (EDA) based on 38 measurement runs collected over ten months along a rural 64 km route in Germany. The dataset includes passive mobile network signal quality parameters, Global Navigation Satellite System (GNSS) position and precision data, as well as contextual features, such as speed, driving direction, day of the week, weather, and distance to the connected base station. Although mean values capture overall tendencies for areas with consistently good or poor coverage, they fail to capture the variability necessary for reliable prediction on a per-trip basis. Notably, some route segments show high variance in signal quality across different measurement runs. This variability is assumed to result from changing environmental influences, such as weather or traffic conditions at different times. Our analysis reveals weak but statistically relevant correlations between several contextual features (e.g., temperature ≈ -0.2) and network quality indicators. The inclusion of weather parameters or the day of the week has been shown to lower the Mean Absolute Error (MAE) compared to a prediction based only on measurements from the past. These findings underscore the importance of contextual information and localized modeling to predict network availability for safety-critical systems, such as autonomous vehicles.
Digitalization in Industry and Society is progressing quickly. Up to now, just 5 static and standard lighting applications have been dominating in the U.S. (Low/High Beam, Daytime Running Lamp Turn Indicator & Position Lamp).
The global interest in standardization of light driven messages from autonomous vehicles to other traffic participants has opened new research needs and research findings. In Europe, GTB has established a working group dedicated to this topic.
The article will discuss the possible contribution of signalling and lighting functions to Digitalization and Autonomous driving by explaining the first elements of functional definition and research results.
The development of machine learning (ML) applications in deep drawing is hindered by limited data availability and the absence of open-access benchmarks for validating novel approaches, including domain generalization over distinct geometries. This paper addresses these challenges by introducing a comprehensive U-shaped dataset tailored to this manufacturing process. Our U-Channel sheet metal (UCSM) dataset combines 90 real-world meshes with an infinite number of synthetic geometry samples generated from four parametric Computer-Aided Design (CAD) models, ensuring extensive geometry variety and data quantity. Additionally, a ready-to-use dataset for drawability assessment and segmentation is provided. Leveraging CAD and mesh data sources bridges the gap between sparse data availability and ML requirements.
Our analysis demonstrates that the proposed parametric models are geometrically valid, and real-world and synthetic data complement each other effectively, providing robust support for ML model development. While the dataset is confined to U-shaped, thin-walled, deep drawing scenarios, it considerably aids in overcoming data scarcity. Thereby, it facilitates the validation and comparison of new geometry-generalizing ML methodologies in this domain. By providing this benchmark dataset, we enhance the comparability and validation of emerging methods for ML advancements in sheet metal forming.
Conversational agents such as Microsoft Copilot and Google Gemini assist users with complex search tasks but often generate misleading or fabricated references. This undermines trust, particularly in high-stakes domains such as medicine and finance. Explainable information retrieval (XIR) aims to address this by making search results more transparent and interpretable. While most XIR research is domain-agnostic, this paper focuses on auditing -- a critical yet underexplored area. We argue that XIR systems can support auditors in completing their complex task. We outline key challenges and future research directions to advance XIR in this domain.
Tackling fake images in cybersecurity - interpretation of a StyleGAN and lifting its black-box
(2025)
In today's digital age, concerns about the dangers of AI-generated images are increasingly common. One powerful tool in this domain is StyleGAN (style-based generative adversarial networks), a generative adversarial network capable of producing highly realistic synthetic faces. To gain a deeper understanding of how such a model operates, this work focuses on analyzing the inner workings of StyleGAN's generator component. Key architectural elements and techniques, such as the Equalized Learning Rate, are explored in detail to shed light on the model's behavior. A StyleGAN model is trained using the PyTorch framework, enabling direct inspection of its learned weights. Through pruning, it is revealed that a significant number of these weights can be removed without drastically affecting the output, leading to reduced computational requirements. Moreover, the role of the latent vector -- which heavily influences the appearance of the generated faces -- is closely examined. Global alterations to this vector primarily affect aspects like color tones, while targeted changes to individual dimensions allow for precise manipulation of specific facial features. This ability to finetune visual traits is not only of academic interest but also highlights a serious ethical concern: the potential misuse of such technology. Malicious actors could exploit this capability to fabricate convincing fake identities, posing significant risks in the context of digital deception and cybercrime.
Autonomous driving, child presence detection, and digital car keys rely on precise localization. Therefore, accurate simulation environments are essential for designing radio systems and optimizing their positioning algorithms. A critical aspect of developing a reliable localization system is modeling the physical radio channel, which includes the transmitting antenna, the propagation medium, and the receiving antenna. However, existing channel models often overlook important physical effects of antennas, such as frequency dependence and polarization, leading to discrepancies between simulations and real-world measurements. To address this, the paper enhances full polarimetric channel models by incorporating the frequency dependence of both antennas and propagation channels. The theoretical model is validated through Vector Network Analyzer channel sounding measurements using reference antennas. This approach provides a foundation for polarization-based post-processing, ultimately improving positioning accuracy.
Significant advances in using conductive textiles have been made over the past two decades, particularly in knitted fabrics. The aim is to deepen understanding of how knitted structures, especially those with hybrid yarns containing fine conductive elements (microwires), interact with electromagnetic fields. To establish simulation models that align with subsequent measurements, it is crucial to determine the dielectric properties, specifically relative permittivity and loss tangent/conductivity. This research investigates whether knitted structures with hybrid yarns exhibit anisotropic permittivity, meaning different dielectric properties along orthogonal axes within the fabric plane. These properties were measured using a waveguide technique’ whereby the structures were rotated by 90° to observe changes. The results demonstrate that the knitted structure exhibits metallic-like reflective behavior in one direction, while rotation leads to dielectric material characteristics. This effect is influenced by the number of conductive microwires and the materials used for these wires, such as silver-plated copper or bronze.
Advanced silicon electrode arrays facilitate the recording of thousands of neurons, generating extensive neural data that imposes a significant burden on researchers and processing algorithms. Thus, real-time analysis pipelines are gaining increasing value, while at the same time having to deal with non-stationary and noisy data. We intend to apply Machine Learning (ML) algorithms to a dense set of recordings from rat brains in order to prepare a functional atlas, correlating neuronal signals with anatomical position. While doing so, we needed to decide on a rational way which recording snippet length would best represent the original, longer source sequence and thus suffices to be further processed for anatomical correlation. We implemented an algorithm to evaluate the spectral information of systematically length varied records based on similarity to the original record. For our dataset a recording duration of 3 seconds satisfied moderate requirements across all channels, thus allowing us to reduce computational load for ongoing ML classification of microprobe sourced electrophysiologic signals.