@masterthesis{Ruff2026, type = {Bachelor Thesis}, author = {Ruff, Daniel}, title = {Analyse der psychoakustischen Gr{\"o}ßen von UX Sounds}, publisher = {Technische Hochschule N{\"u}rnberg}, address = {N{\"u}rnberg}, doi = {10.34646/thn/ohmdok-5121}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:92-opus4-51215}, school = {Technische Hochschule N{\"u}rnberg Georg Simon Ohm}, pages = {VIII, 95 S.}, year = {2026}, abstract = {UX-Sounds l{\"o}sen neben ihrer informativen Funktion auch emotionale Reaktionen aus. Die Arbeit untersucht, welche psychoakustischen Gr{\"o}ßen mit bestehenden Valenz/Arousal-Bewertungen von 85 UX-Sounds zusammenh{\"a}ngen, die aus f{\"u}nf studentischen Vorarbeiten zusammengef{\"u}hrt wurden. F{\"u}r jeden Sound wurden sechs psychoakustische Gr{\"o}ßen in MATLAB berechnet (Lautheit, Sch{\"a}rfe, Rauigkeit, Schwankungsst{\"a}rke, Tonalit{\"a}t, Spektralschwerpunkt) und mittels Pearson-Korrelationen mit den V/A-Bewertungen in Beziehung gesetzt. Als stabilster Befund zeigt sich: Ein h{\"o}herer Spektralschwerpunkt geht mit h{\"o}herem Arousal und niedrigerer Valenz einher - {\"u}ber alle Teilgruppen hinweg. Andere Zusammenh{\"a}nge, etwa Tonalit{\"a}t-Arousal, erwiesen sich hingegen als stark quellenabh{\"a}ngig, was die Bedeutung der Datensatzzusammensetzung unterstreicht.}, subject = {Psychoakustik}, language = {de} } @incollection{Walther2026, author = {Walther, Christoph}, title = {Personzentrierter Ansatz}, series = {W{\"o}rterbuch Soziale Arbeit}, volume = {2026}, booktitle = {W{\"o}rterbuch Soziale Arbeit}, editor = {Amthor, Ralph-Christian and Goldberg, Brigitta and Hansbauer, Peter and Landes, Benjamin and Wintergerst, Theresia}, edition = {10. Aufl.}, publisher = {Beltz Juventa}, address = {Weinheim}, isbn = {978-3-7799-9000-0}, pages = {4}, year = {2026}, abstract = {Stichwort im W{\"o}rterbuch der Sozialen Arbeit}, subject = {Psychologie}, language = {de} } @incollection{Walther2026, author = {Walther, Christoph}, title = {Soziale Arbeit mit psychisch erkrankten Menschen}, series = {W{\"o}rterbuch Soziale Arbeit}, volume = {2026}, booktitle = {W{\"o}rterbuch Soziale Arbeit}, editor = {Amthor, Ralph-Christian and Goldberg, Brigitta and Hansbauer, Peter and Landes, Benjamin and Wintergerst, Theresia}, edition = {10. Aufl.}, publisher = {Beltz Juventa}, address = {Weinheim}, isbn = {978-3-7799-9000-0}, pages = {5}, year = {2026}, abstract = {Stichwort "Soziale Arbeit mit psychisch erkrankten Menschen" im W{\"o}rterbuch Soziale Arbeit}, language = {de} } @misc{Walther2025, author = {Walther, Christoph}, title = {50 Jahre Psychiatrie-Enquete 1971-1975. Hintergr{\"u}nde - Positionen - Reformvorschl{\"a}ge}, year = {2025}, abstract = {50 Jahre Psychiatrie-Enquete 1975: Politische und fachliche Hintergr{\"u}nde zu der Entstehung. Erarbeitete Positionen, Forderungen zur Ver{\"a}nderung und konkrete Reformvorschl{\"a}ge in der psychiatrischen Versorgung der BRD in der Enquete.}, language = {de} } @misc{Walther2026, author = {Walther, Christoph}, title = {Mentale Gesundheit}, volume = {2026}, year = {2026}, abstract = {Was versteht man unter mentaler Gesundheit? Woran kann man sie erkennen? Woran kann man erkennen, dass sie gef{\"a}hrdet ist?}, subject = {Psychologie}, language = {de} } @inproceedings{YoendemTavakoliKolagariSchlerethGroh2025, author = {Y{\"o}ndem, Şakir Furkan and Tavakoli Kolagari, Ramin and Schlereth-Groh, Benedikt}, title = {Bridging the Sim-to-Real Gap with Explainability for ML-based Object Detection on Sonar Data}, pages = {8}, year = {2025}, abstract = {Saving drowning victims is time-critical, but detecting people underwater is highly challenging due to poor visibility and large distances. While side-scan sonar (SSS) is widely used for seafloor mapping and debris detection, human detection in sonar data remains largely unexplored. Training deep neural networks for this task requires a large dataset, but collecting real maritime data is difficult and expensive, making a synthetic data generation approach necessary. We introduce SimWave, a simulation environment designed to generate synthetic data for underwater human detection. We train deep learning models on real, synthetic, and hybrid datasets, evaluating their performance on real sonar images. The contribution of this paper lies in combining synthetic data generation with Explainable Artificial Intelligence (XAI) to systematically refine artificial datasets, addressing the gap between synthetic and natural data to enhance real-world performance—-an approach not previously explored in underwater sonar-based human detection. To gain insight into the model's decision-making process, we apply XAI techniques to analyze how attention shifts between real and synthetic training data. This helps visualize the synthetic-real data mismatch, refine synthetic data, and enhance model performance in real-world conditions. Our experimental results show that models trained on hybrid datasets, supported by XAI-based analysis, achieve notable performance improvements and better generalization. XAI helps identify domain gaps between real and synthetic data, allowing for dataset refinement and improved model accuracy. These findings highlight the effectiveness of synthetic generated data in training deep learning models for underwater human detection and emphasize the critical role of XAI in optimizing training data for real-world conditions}, language = {en} } @inproceedings{YoendemSchlerethGrohTavakoliKolagari2026, author = {Y{\"o}ndem, Şakir Furkan and Schlereth-Groh, Benedikt and Tavakoli Kolagari, Ramin}, title = {MirrorDrive: A Sim-to-Real Bridge for Reproducible Automotive Software Development}, publisher = {Gesellschaft f{\"u}r Informatik}, address = {Bonn}, doi = {10.18420/se2026-ws_02}, pages = {12}, year = {2026}, abstract = {A central difficulty in automotive software engineering is to ensure that autonomous driving functions are verified and validated with equal rigor in simulation and on physical vehicles. The transition from virtual development to real-world deployment introduces domain differences in sensor noise, actuator behavior, latency, timing, and environmental conditions, which complicate reproducibility and challenge the reliability of sim-to-real transfer. To address these issues, we present MirrorDrive, an ROS 2-based platform designed to enforce a "zero code-change" principle. MirrorDrive uses a 1:1 Gazebo simulation model of a physical remote-controlled (RC)-Car platform, allowing the exact same application code to run without adaptation in both environments, enabled by the ROS 2 middleware. The core contribution lies in achieving full deployment consistency by packaging the entire environment and application code within Docker containers. This approach guarantees reproducibility across different systems, supporting Continuous Integration (CI). Our case studies suggest that MirrorDrive provides a scalable and portable framework for the rapid prototyping and reliable validation of autonomous software, making it a valuable testbed for sim-to-real research and education (https://github.com/furkanyondemm/MirrorDrive).}, language = {en} } @inproceedings{YoendemSchlerethGrohKolagari2026, author = {Y{\"o}ndem, Sakir Furkan and Schlereth-Groh, Benedikt and Kolagari, Ramin Tavakoli}, title = {AquaScan-1K: Paving the Way to Robust Object Detection in Side-Scan Sonar}, series = {2026 IEEE Canadian Atlantic Ocean Symposium (CAOS)}, booktitle = {2026 IEEE Canadian Atlantic Ocean Symposium (CAOS)}, publisher = {IEEE}, doi = {10.1109/CAOS69069.2026.11665924}, pages = {1 -- 6}, year = {2026}, abstract = {Machine learning models for underwater perception face substantial challenges in turbid search-and-rescue scenarios, where optical sensing is ineffective and side-scan sonar (SSS) becomes the primary modality. Existing SSS datasets often lack variability in viewpoint and seabed structure, limiting robustness and generalization. We present AquaScan-1K, a curated dataset of approximately 1,000 SSS images acquired under multiple angles and diverse seabed conditions, with expert annotations provided by Red Cross personnel using ethical human surrogates. The dataset captures realistic operational variability relevant for small-object detection. Cross-dataset experiments show that models trained on AquaScan-1K generalize more reliably under domain shift than those trained on public baselines. Transformer architectures exhibit reduced performance degradation under domain shift, and saliency analyses confirm that they consistently attend to shadow-silhouette features that constitute the primary discriminative signal in side-scan sonar imagery.}, language = {en} } @inproceedings{YoendemSchlerethGrohTavakoliKolagari2026, author = {Y{\"o}ndem, Sakir Furkan and Schlereth-Groh, Benedikt and Tavakoli Kolagari, Ramin}, title = {WaveGuard: Real-Time On-Board Detection of Missing Persons in Side-Scan Sonar Imagery}, year = {2026}, language = {en} } @incollection{SchlerethGrohYoendemKolagarietal.2026, author = {Schlereth-Groh, Benedikt and Y{\"o}ndem, Şakir Furkan and Kolagari, Ramin Tavakoli and Schmid, Ute}, title = {Explainable-AI-Based Training for Relevance-Based Robust Reinforcement Learning}, series = {Lecture Notes in Computer Science}, booktitle = {Lecture Notes in Computer Science}, publisher = {Springer Nature Switzerland}, address = {Cham}, isbn = {9783032323347}, issn = {0302-9743}, doi = {10.1007/978-3-032-32335-4_28}, pages = {309 -- 316}, year = {2026}, language = {en} }