OPEN FAU Online-Publikationssystem der Friedrich-Alexander-Universität Erlangen-Nürnberg
Not a member yet
23478 research outputs found
Sort by
Gepulste neuronale Netze für die Klassifikation ereignisgesteuerter FMCW-Radarsignale
Get PDFIn recent years, the expansion of Internet of Things (IoT) devices has been evident in various applications, including smart homes, automotive interfaces, energy, and transportation. The appeal of these devices lies in their broad applicability, simplicity, user-friendly experience, and the healthier environment they foster. The desirability of wireless interaction with these devices is unquestionable. This results in the convenience of wirelessly controlling a smart TV, adjusting the volume, switching it on or off, and managing an air conditioner’s temperature or power status without physical contact. Even vending machines could benefit from this technology, offering touch-free options such as obtaining a transportation ticket wirelessly, a handy feature in the wake of widespread viruses and diseases like the recent pandemic. All these wireless interactions could be facilitated through hand motion detection or Hand Gesture Recognition (HGR).
Several technologies, such as cameras, enable wireless interaction with smart devices. However, the limitations of cameras - their size, cost, privacy concerns, and sensitivity to weather and lighting – call for alternative solutions. Radar technology offers a compelling alternative: compact, privacy-preserving, weather and lighting-resistant, and cost-effective. Research efforts have focused on harnessing the power of Artificial Neural Networks (ANNs) with radar signals for efficient and accurate smart device control. However, pursuing extreme-edge solutions has increased interest in Spiking Neural Networks (SNNs). SNNs, drawing inspiration from biological systems, exhibit superior biological plausibility and capture complex dynamics more effectively. Their communication via discrete electrical impulses (spikes) results in sparse communication between layers, further enhanced by the single-bit nature of these spikes. This translates to replacing Multiply-and-Accumulate (MAC) arrays with simpler adders, significantly reducing computational complexity and boosting energy efficiency.
This work investigates SNNs as a key to unlocking energy-efficient HGR solutions using radar technology. We take advantage of the biological plausibility of spiking neurons within SNNs to enable HGR directly from time-domain radar data. Furthermore, we employ a specific type of spiking neuron – the Resonate-and-Fire (RAF) neuron – capable of efficiently converting time-domain data into spatio-temporal spikes. These spikes represent and encode the fundamental frequencies within the radar signals, paving the way for robust HGR. All proposed approaches are evaluated using real-world data. Our results demonstrate that SNN-based solutions achieve performance comparable to traditional methods. These findings prove that SNNs can be exploited to improve HGR by directly processing radar sensor signals. Moreover, the inherent sparsity of spiking neurons motivates the exploration of neuromorphic hardware implementations, offering substantial energy efficiency benefits for resource-constrained systems.In den letzten Jahren hat die Verbreitung von IoT-Systemen in verschiedenen Anwendungsbereichen wie Smart Home, Automotiveschnittstellen, Energie und Transportwesen deutlich zugenommen. Die Attraktivität dieser Geräte liegt in ihrer breiten Anwendbarkeit, Einfachheit, Benutzerfreundlichkeit und der damit verbundenen Verbesserung der Umweltbedingungen. Es ist unumstritten, dass eine kontaktlose Bedienung solcher Geräte wünschenswert ist. Beispielsweise lässt sich so ein Smart-TV bequem drahtlos steuern, die Lautstärke regeln, das Gerät ein- oder ausschalten, oder bei einer Klimaanlage kann die Temperatur und der Energiestatus ohne physischen Kontakt abgefragt werden. Darüber hinaus könnten auch Verkaufsautomaten von dieser Technologie profitieren, indem sie kontaktlose Optionen anbieten, wie z. B. das drahtlose Entwerten eines Fahrscheins, was besonders bei Pandemien wie der kürzlichen Covid-19-Pandemie eine nützliche Funktion darstellen würde. All diese drahtlosen Interaktionen könnten durch die Erkennung von Handbewegungen oder Handgesten erleichtert werden.
Verschiedene Technologien, wie z. B. Kameratechniken, ermöglichen derzeit die kontaktlose Interaktion mit Smart Devices. Jedoch zeigen Kameras Einschränkungen, was Größe, Kosten, Datenschutzbedenken und Empfindlichkeit gegenüber Wetter und Licht angeht, sodass alternative Lösungen erforderlich sind. Für die genannten Punkte bietet die Radartechnologie eine attraktive Alternative: Sie ist kompakt, wahrt die Privatsphäre, ist unabhängig von Wetter und Beleuchtung und kostengünstig. Die Forschung hat sich darauf konzentriert, die Fähigkeiten von künstlichen neuronalen Netzen (ANNs) in Kombination mit Radar für eine effiziente und genaue Steuerung intelligenter Geräte zu nutzen. Das Streben nach portablen Lösungen, die in Edge-Systemen eingebettet sind, hat dabei das Interesse an gepulsten neuronalen Netzen (SNNs) erhöht. Bioinspirierte SNNs weisen eine höhere Plausibilität auf und erfassen komplexe dynamische Szenarien effektiver. Ihre Kommunikation über diskrete elektrische Impulse (Spikes) führt zu einer spärlichen Kommunikation zwischen den Schichten, die durch die Ein-Bit-Natur dieser Spikes noch verstärkt wird. Dadurch können MAC-Arrays durch einfachere Addierer ersetzt werden, was die Rechenkomplexität drastisch reduziert und die Energieeffizienz erhöht.
In dieser Arbeit werden SNNs als Schlüssel zur Realisierung energieeffizienter HGR-Lösungen mit Hilfe der Radartechnologie untersucht. Es wird die biologie-gestützte Plausibilität von feuernden Neuronen innerhalb von SNNs ausgenutzt, um die HGR direkt aus Radardaten im Zeitbereich ableiten zu können. Darüber hinaus wird ein spezieller Typ von Spiking-Neuronen – das RAF-Neuron – verwendet, welches in der Lage ist, Zeitdaten effizient in räumlich-zeitliche Spikes umzuwandeln. Diese Spikes kodieren die Grundfrequenzen der Radarsignale und ermöglichen so eine robuste HGR. Alle vorgeschlagenen Ansätze werden anhand realer Daten evaluiert. Die Ergebnisse zeigen, dass SNN-basierte Lösungen eine vergleichbare Performance erzielen wie Lösungen, die auf herkömmlichen Methoden basieren. Diese Arbeit zeigt auf, dass SNNs zur Verbesserung der HGR genutzt werden können, indem Radarsensorsignale direkt verarbeitet werden. Darüber hinaus motiviert die Spärlichkeit von Spike-Neuronen die Erforschung neuromorpher Hardware-Implementierungen, die erhebliche Energieeffizienzvorteile für ressourcenbeschränkte Systeme bieten
A genome-wide association study with tissue transcriptomics identifies genetic drivers for classic bladder exstrophy
Get PDFClassic bladder exstrophy represents the most severe end of all human congenital anomalies of the kidney and urinary tract and is associated with bladder cancer susceptibility. Previous genetic studies identified one locus to be involved in classic bladder exstrophy, but were limited to a restrict number of cohort. Here we show the largest classic bladder exstrophy genome-wide association analysis to date where we identify eight genome-wide significant loci, seven of which are novel. In these regions reside ten coding and four non-coding genes. Among the coding genes is EFNA1, strongly expressed in mouse embryonic genital tubercle, urethra, and primitive bladder. Re-sequence of EFNA1 in the investigated classic bladder exstrophy cohort of our study displays an enrichment of rare protein altering variants. We show that all coding genes are expressed and/or significantly regulated in both mouse and human embryonic developmental bladder stages. Furthermore, nine of the coding genes residing in the regions of genome-wide significance are differentially expressed in bladder cancers. Our data suggest genetic drivers for classic bladder exstrophy, as well as a possible role for these drivers to relevant bladder cancer susceptibility.A genome-wide association study on classic bladder exstrophy reveals eight genome-wide significant loci, most of which contained genes expressed in embryonic developmental bladder stages.Deutsche Forschungsgemeinschaft (German Research Foundation) https://doi.org/10.13039/50110000165
Leveraging Large Language Models to Generate Course‐Specific Semantically Annotated Learning Objects
Get PDFABSTRACT Background Over the past few decades, the process and methodology of automatic question generation (AQG) have undergone significant transformations. Recent progress in generative natural language models has opened up new potential in the generation of educational content. Objectives This paper explores the potential of large language models (LLMs) for generating computer science questions that are sufficiently annotated for automatic learner model updates, are fully situated in the context of a particular course and address the cognitive dimension understand . Methods Unlike previous attempts that might use basic methods such as ChatGPT, our approach involves more targeted strategies such as retrieval‐augmented generation (RAG) to produce contextually relevant and pedagogically meaningful learning objects. Results and Conclusions Our results show that generating structural, semantic annotations works well. However, this success was not reflected in the case of relational annotations. The quality of the generated questions often did not meet educational standards, highlighting that although LLMs can contribute to the pool of learning materials, their current level of performance requires significant human intervention to refine and validate the generated content.Federal Ministry of Education and Research German
Selected annotated instance segmentation sub-volumes from a large scale CT data-set of a historic aircraft
Get PDFThe Me 163 was a Second World War fighter airplane and is currently displayed in the Deutsches Museum in Munich, Germany. A complete computed tomography (CT) scan was obtained using a large scale industrial CT scanner to gain insights into its history, design, and state of preservation. The CT data enables visual examination of the airplane’s structural details across multiple scales, from the entire fuselage to individual sprockets and rivets. However, further processing requires instance segmentation of the CT data-set. Currently, there are no adequate computer-assisted tools for automated or semi-automated segmentation of such large scale CT airplane data. As a first step, an interactive data annotation process has been established. So far, seven 512 × 512 × 512 voxel sub-volumes of the Me 163 airplane have been annotated, which can potentially be used for various applications in digital heritage, non-destructive testing, or machine learning. This work describes the data acquisition process, outlines the interactive segmentation and post-processing, and discusses the challenges associated with interpreting and handling the annotated data.This work was supported by the Bavarian Ministry of Economic Affairs, Regional Development and Energy through the Center for Analytics Data Applications (ADA-Center) within the framework of, BAYERN DIGITAL II`` (20-3410-2-9-8
Γ-Convergence and Stochastic Homogenization of Second-Order Singular Perturbation Models for Phase Transitions
No full textWe study the effective behavior of random, heterogeneous, anisotropic, second-order phase transitions energies that arise in the study of pattern formations in physical–chemical systems. Specifically, we study the asymptotic behavior, as εgoes to zero, of random heterogeneous anisotropic functionals in which the second-order perturbation competes not only with a double well potential but also with a possibly negative contribution given by the first-order term. We prove that, under suitable growth conditions and under a stationarity assumption, the functionals Γ-converge almost surely to a surface energy whose density is independent of the space variable. Furthermore, we show that the limit surface density can be described via a suitable cell formula and is deterministic when ergodicity is assumed.Open Access funding enabled and organized by Projekt DEAL.Deutsche Forschungsgemeinschafthttp://dx.doi.org/10.13039/501100001659Friedrich-Alexander-Universität Erlangen-Nürnberg (1041
Prior flavivirus immunity skews the yellow fever vaccine response to cross-reactive antibodies with potential to enhance dengue virus infection
Get PDFThe yellow fever 17D vaccine (YF17D) is highly effective but is frequently administered to individuals with pre-existing cross-reactive immunity, potentially impacting their immune responses. Here, we investigate the impact of pre-existing flavivirus immunity induced by the tick-borne encephalitis virus (TBEV) vaccine on the response to YF17D vaccination in 250 individuals up to 28 days post-vaccination (pv) and 22 individuals sampled one-year pv. Our findings indicate that previous TBEV vaccination does not affect the early IgM-driven neutralizing response to YF17D. However, pre-vaccination sera enhance YF17D virus infection in vitro via antibody-dependent enhancement (ADE). Following YF17D vaccination, TBEV-pre-vaccinated individuals develop high amounts of cross-reactive IgG antibodies with poor neutralizing capacity. In contrast, TBEV-unvaccinated individuals elicit a non-cross-reacting neutralizing response. Using YF17D envelope protein mutants displaying different epitopes, we identify quaternary dimeric epitopes as the primary target of neutralizing antibodies. Additionally, TBEV-pre-vaccination skews the IgG response towards the pan-flavivirus fusion loop epitope (FLE), capable of mediating ADE of dengue and Zika virus infections in vitro. Together, we propose that YF17D vaccination conceals the FLE in individuals without prior flavivirus exposure but favors a cross-reactive IgG response in TBEV-pre-vaccinated recipients directed to the FLE with potential to enhance dengue virus infection.Flavivirus infection or vaccination can induce cross-reactive immune responses. Here, the authors show how previous immunization with the tick-borne encephalitis virus vaccine affects the immune response to the yellow fever vaccine, suggesting that the yellow fever vaccine virus conceals epitopes shared with other flaviviruses in flavivirus-naive but not flavivirus-pre-exposed individuals.Deutsche Forschungsgemeinschaft (German Research Foundation)https://doi.org/10.13039/501100001659EC | Horizon 2020 Framework Programme (EU Framework Programme for Research and Innovation H2020)https://doi.org/10.13039/100010661Agence Nationale de la Recherche (French National Research Agency)https://doi.org/10.13039/501100001665Friedrich-Baur-Stiftung (Friedrich Baur Stiftung)https://doi.org/10.13039/501100008426Bundesministerium für Bildung und Forschung (Federal Ministry of Education and Research)https://doi.org/10.13039/50110000234
Square-stepping exercise in older inpatients in early geriatric rehabilitation. A randomized controlled pilot study
Get PDFBackground Preservation of mobility and fall prevention have a high priority in geriatric rehabilitation. Square-Stepping Exercise (SSE) as an evaluated and standardized program has been proven to be an effective training for older people in the community setting to reduce falls and improve subjectively perceived health status. This randomized controlled trial (RCT), for the first time, examines SSE in the context of inpatient early geriatric rehabilitation compared to conventional physiotherapy (cPT). Methods Data were collected in a general hospital in the department of acute geriatric care at admission and discharge. Fifty-eight inpatients were randomized to control (CG, n = 29) or intervention groups (IG, n = 29). CG received usual care with cPT five days per week during their hospital stay. For the IG SSE replaced cPT for at least six sessions, alternating with cPT. Physical function was measured with the Short Physical Performance Battery (SPPB) and Timed “Up & Go” (TUG). Gait speed was measured over a distance of 10 m. In a subgroup ( n = 17) spatiotemporal gait parameters were analyzed via a GAITRite® system. Results Both the SPPB total score improved significantly ( p = < 0.001) from baseline to discharge in both groups, as did the TUG ( p < 0.001). In the SPPB Chair Rise both groups improved with a significant group difference in favor of the IG ( p = 0.031). For both groups gait characteristics improved: Gait speed ( p = < 0.001), walk ratio ( p = 0.011), step length ( p = < 0.001), stride length ( p = < 0.001) and double support ( p = 0.009). For step length at maximum gait speed ( p = 0.054) and stride length at maximum gait speed ( p = 0.060) a trend in favor of the IG was visible. Conclusions SSE in combination with a reduced number of sessions of cPT is as effective as cPT for inpatients in early geriatric rehabilitation to increase physical function and gait characteristics. In the Chair Rise test SSE appears to be superior. These results highlight that SSE is effective, and may serve as an additional component for cPT for older adults requiring geriatric acute care. Trial Registration DRKS00026191.Open Access funding enabled and organized by Projekt DEAL.Carl von Ossietzky Universität Oldenburg (3092
The population context is a driver of the heterogeneous response of epithelial cells to interferons
Get PDFIsogenic cells respond in a heterogeneous manner to interferon. Using a micropatterning approach combined with high-content imaging and spatial analyses, we characterized how the population context (position of a cell with respect to neighboring cells) of epithelial cells affects their response to interferons. We identified that cells at the edge of cellular colonies are more responsive than cells embedded within colonies. We determined that this spatial heterogeneity in interferon response resulted from the polarized basolateral interferon receptor distribution, making cells located in the center of cellular colonies less responsive to ectopic interferon stimulation. This was conserved across cell lines and primary cells originating from epithelial tissues. Importantly, cells embedded within cellular colonies were not protected from viral infection by apical interferon treatment, demonstrating that the population context-driven heterogeneous response to interferon influences the outcome of viral infection. Our data highlights that the behavior of isolated cells does not directly translate to their behavior in a population, placing the population context as one important factor influencing heterogeneity during interferon response in epithelial cells.Synopsis The heterogeneous response to interferons in epithelial cells is determined by the population context (i.e. a cell’s positions within a population) through cell polarization and basolateral localization of the interferon receptor. Isogenic epithelial cells respond in a heterogeneous manner to type I and type III interferons. The population context (position of a cell within a population) drives the heterogeneous response to interferons, with cells located at the edge of a population being more responsive than cells located in the center of a population. Cells in the center of a population are refractive to interferon treatment from their apical side due to the polarized basolateral distribution of the interferon receptors. By controlling the response to interferon treatment, the population context affects susceptibility to viral infection.The heterogeneous response to interferons in epithelial cells is determined by the population context (i.e. a cell’s positions within a population) through cell polarization and basolateral localization of the interferon receptor.Deutsche Forschungsgemeinschaft (DFG)http://dx.doi.org/10.13039/501100001659Chica and Heinz Schaller Foundation (CHS-Stiftung)http://dx.doi.org/10.13039/501100012284Deutscher Akademischer Austauschdienst (DAAD)http://dx.doi.org/10.13039/501100001655UF | UF Health | College of Medicine, University of Florida (UF College of Medicine)http://dx.doi.org/10.13039/10000827
Inelastic neutron scattering: A unique tool to study hydrogen in materials
Get PDFInelastic neutron scattering (INS) spectroscopy can measure the vibrational spectra of materials on the whole range of vibrational motions (0–4400 cm -1) and effectively open up the field of neutron spectroscopy. Unlike optical spectroscopy, INS is a technique mainly used to study hydrogen-containing materials due to the high cross section of hydrogen and the lack of selection rules. The VISION spectrometer at the SNS in Oak Ridge National Laboratory has a flux at low energy transfers up to 40 times over its predecessors and has unprecedented sensitivity. Neutrons penetrate easily through materials, metals in particular, making complex sample environments and gas dosing relatively easy. Comparison of the INS spectra with computer models is rigorous and necessary to interpret INS data. This paper presents examples of the technique’s unique capabilities for studying metal hydrides and molecular hydrogen in confinement. Graphical abstrac