Refine
Year of publication
Document Type
- Article (1816)
- conference proceeding (article) (1678)
- Part of a Book (565)
- conference proceeding (presentation, abstract) (263)
- Book (160)
- conference talk (137)
- Report (122)
- Review (100)
- Working Paper (49)
- Preprint (47)
- conference proceeding (volume) (34)
- Doctoral Thesis (25)
- Edited collection (21)
- Part of Periodical (20)
- Master's Thesis (15)
- Lecture (12)
- Bachelor-/Diplom Thesis (10)
- Patent (10)
- Other (9)
- Moving Images (7)
- Study Thesis (4)
- Sound (3)
- Course Material (2)
- Habilitation (2)
Language
Is part of the Bibliography
- no (5111)
Keywords
- Simulation (37)
- Digitalisierung (29)
- Erneuerbare Energien (28)
- Energiewende (27)
- field emission (24)
- Bildgebendes Verfahren (22)
- Deutschland (22)
- Klimaneutralität (22)
- Deep Learning (21)
- Power-to-Gas (21)
Institute
- Fakultät Angewandte Sozial- und Gesundheitswissenschaften (1292)
- Fakultät Informatik und Mathematik (1130)
- Fakultät Maschinenbau (893)
- Fakultät Elektro- und Informationstechnik (796)
- Institut für Sozialforschung und Technikfolgenabschätzung (IST) (552)
- Labor Empirische Sozialforschung (458)
- Fakultät Angewandte Natur- und Kulturwissenschaften (403)
- Fakultät Bauingenieurwesen (351)
- Labor für Technikfolgenabschätzung und Angewandte Ethik (LaTe) (310)
- Laboratory for Safe and Secure Systems (LAS3) (277)
Begutachtungsstatus
- peer-reviewed (1552)
- begutachtet (92)
The present study investigates the interface between carbon steel and titanium samples annealed at different temperatures (𝜗1 = 800 ◦C and 𝜗2 = 1050 ◦C). In both cases, an observable layer forms at the interface, with its thickness increasing from t𝜗1= 2.75 ± 0.15 μm at 800 ◦C to t𝜗2= 8.86 ± 0.29 μm at 1050 ◦C. The layer’s composition and thickness evolve with temperature. Analysis reveals approximately 40 at.-% carbon concentration in the exterior region, indicating likely titanium carbide creation. X-ray diffraction identifies titanium carbide peaks, while microscopy and elemental mapping confirm compositional gradients at the interface.
Electron Backscatter Diffraction (EBSD) shows a gradient in grain size near the TiC surface, reflecting TiC nucleation rates. XRD data detect both titanium carbide and titanium phases, with TiC becoming more prominent at 1050 ◦C. Rietveld analysis further confirms TiC formation. Notably, distinct diffraction patterns on the contact and rear sides suggest a Ti(C, O, N) presence. Depth profiles exhibit varying surface and depth carbon concentrations, attributed to temperature effects. The study successfully demonstrates TiC coating fabrication through hot pressing, wherein Ti(C, O, N) coatings arise from titanium’s affinity for reacting with oxygen and nitrogen. This research contributes to the understanding of phase transformations and interfacial properties in titanium-carbon steel systems.
This article describes a contactless fiber-optic position sensor. It comprises a Polymer Optical Fiber that is grinded to form a D-shaped cross-section with an exposed fiber core. This sensor has two photodiodes at both fiber ends to measure the emitted light intensity. Light is coupled using a red LED at the side face into the exposed core of the fiber at a defined position. The position of the LED at the length of the fiber is measured by calculating the optical power quotient measured by both photodiodes. To test this sensor, Polymer Optical Fibers with different side surface roughness are produced and qualified. Measurements show that the optical power quotient is reproducible and nearly linear over the length of the fiber. It is also seen that the fiber attenuation increases when grinding the fiber side-face with rougher sandpaper. Position measurements show an absolute position error of this sensor in the range of a few millimeters. Microscope images show surface defects along the polished side face of the fiber that are expected to lead to a nonuniform attenuation along the fiber and cause the position errors. Overall, it is proven that this sensor principle works as a contactless low-cost position sensor for short distances with an absolute position standard deviation error lower than 1 mm.
In this work, high-current field emission electron source chips were fabricated using laser-micromachining and MEMS technology. The resulting chips were combined with commercially available printed circuit boards (PCBs) to obtain a multichip electron source. By controlling the separate electron sources using an external current control circuit, we were able to divide the desired total current evenly across the individual chips deployed in the PCB-carrier. In consequence, we were able to show a decreased degradation due to the reduced current load per chip. First, a single electron source chip was measured without current regulation. A steady-state emission current of 1 mA with a high stability of ±1.3% at an extraction voltage of 250 V was observed. At this current level, a mean degradation slope of −0.7 μA/min with a nearly perfect transmission ratio of 99% ± 0.4% was determined. The measurements of a fully assembled multichip PCB-carrier electron source, using a current control circuit for regulation, showed that an even distribution of the desired total current led to a decreased degradation. This was determined by the increase in the required extraction voltage over time. For this purpose, two current levels were applied to the electron source chips of the PCB-carrier using an external current control circuit. First, 300 μA total current was evenly distributed among the individual electron source chips followed by the emission of 300 μA per electron source chip. This allows the observation of the influence of a distributed and nondistributed total current, carried by the electron source chips. Thereby, we obtained an increase in the mean degradation slope from +0.011 V/min (300 μA distributed) to +0.239 V/min (300 μA per chip), which is approximately 21 times higher. Moreover, our current control circuit improved the current stability to under 0.1% for both current levels, 300 μA distributed and 300 μA per chip.
BACKGROUND
Differentiation of high-flow from low-flow vascular malformations (VMs) is crucial for therapeutic management of this orphan disease.
OBJECTIVE
A convolutional neural network (CNN) was evaluated for differentiation of peripheral vascular malformations (VMs) on T2-weighted short tau inversion recovery (STIR) MRI.
METHODS
527 MRIs (386 low-flow and 141 high-flow VMs) were randomly divided into training, validation and test set for this single-center study. 1) Results of the CNN's diagnostic performance were compared with that of two expert and four junior radiologists. 2) The influence of CNN's prediction on the radiologists' performance and diagnostic certainty was evaluated. 3) Junior radiologists' performance after self-training was compared with that of the CNN.
RESULTS
Compared with the expert radiologists the CNN achieved similar accuracy (92% vs. 97%, p = 0.11), sensitivity (80% vs. 93%, p = 0.16) and specificity (97% vs. 100%, p = 0.50). In comparison to the junior radiologists, the CNN had a higher specificity and accuracy (97% vs. 80%, p < 0.001; 92% vs. 77%, p < 0.001). CNN assistance had no significant influence on their diagnostic performance and certainty. After self-training, the junior radiologists' specificity and accuracy improved and were comparable to that of the CNN.
CONCLUSIONS
Diagnostic performance of the CNN for differentiating high-flow from low-flow VM was comparable to that of expert radiologists. CNN did not significantly improve the simulated daily practice of junior radiologists, self-training was more effective.
In this work we assess whether there is information in pauses in-between utterances of the same or different speakers that are predictive of the following speaker’s utterance. We present models that connect a person’s visual features before they speak to their upcoming utterance. In our experiments we find that outof-the-box pre-trained models can already reach a better-than-chance performance in correlating video embeddings to utterance embeddings. In contrast, models that attempt to predict the first word after the pause do not outperform a unigram model, indicating that our models do not read lips (based e.g. on co-articulation effects) but rather capture more fundamental aspects of the upcoming utterance.
Speech Recognition Errors in ASR Engines and Their Impact on Linguistic Analysis in Psychotherapies
(2024)
Modern intervention planning in psychotherapies may benefit from predicting process relevant psychotherapy constructs by automated speech analysis. One essential step is the extraction of relevant linguistic speech markers by ASR engines, which because of highly sensible data, work offline. We analyze transcription errors from NeMo, Whisper, and Wav2Vec2.0, focusing on their impact on linguistic markers usually requiring high quality transcripts. By utilizing part-of-speech tagging, we examine error occurrences among different word types. The Linguistic Inquiry and Word Count (LIWC) software aids in extracting markers. We highlight challenges in transcribing spontaneous speech, prevalent in therapy, and compare results with the Mozilla CommonVoice dataset, which features read speech.
Despite the potential of AI, only a small percentage of small and medium-sized enterprises (SMEs) are adopting it due to data issues, expertise gaps, and implementation barriers. Zero-shot learning offers a promising approach for SMEs by minimizing these obstacles. This paper explores the use of zero-shot learning in a real-world NLP classification task on online comments (comparable with intent classification tasks) from the e-learning platform Sofatutor. While finetuning has achieved high accuracy (82.3–86.5%), zero-shot models have shown lower performance (39.3–61.4%) due to different label selection, grouping of different scenarios in one class and the type of classification task. Even if the current accuracy is not sufficient for practical application, pre-filtering the data using zeroshot learning might be a promising option for SMEs.
Wine making is usually considered a domain being far off the processing of speech and language. But in a particular aspect, the domains of speech processing and wine making are related, namely, in the description of wine aromas. These descriptors are used for creating wine expertise as well as more general (advertisement-like) textual representations. In the current paper, we use Natural Language Processing techniques, especially Named Entity Recognition, to identify Aspects and Opinions, reflecting wine characteristics. These are combined with analyses of respective relations (triplet extraction) building Aspect-Opinion-Pairs to establish indicative aroma descriptors, also trying to approach the complex interplay amongst these individual statements. In our experiments, we rely on the Falstaff corpus comprising a huge set of wine descriptions. This results in an average F1 score of around 0.85 for Aspect-Opinion classification. For triplet generation multiple strategies were compared, resulting in an average F1 score of 0.67 in this challenging task. For both tasks we rely only on a handful of manually annotated samples, applying pseudo-labeling methods from seed data to achieve automatic labeling.
Recent neural text-to-speech (TTS) models are able to synthesize highly
natural speech signals using deep learning techniques. In practical applications, it can be desirable to have explicit control over the prosody (speech rate, fundamental frequency, and energy) of the synthesized speech. Such controllability can be achieved by adding prosody prediction modules, whose main purpose is to estimate plausible prosody features for each phoneme in the text input. This explicit modeling also allows for changing prosody features at inference time, consequently enabling the adjustment of the prosody in the synthesized audio. In this paper, we evaluate to which extent deliberate manipulation of such prosody features is reflected in the resulting speech audio. We focus particularly on changing the pitch (i.e., fundamental frequency) while applying different normalization strategies.