Dokument-ID Dokumenttyp Verfasser/Autoren Herausgeber Haupttitel Abstract Auflage Verlagsort Verlag Erscheinungsjahr Seitenzahl Schriftenreihe Titel Schriftenreihe Bandzahl ISBN Quelle der Hochschulschrift Konferenzname Quelle:Titel Quelle:Jahrgang Quelle:Heftnummer Quelle:Erste Seite Quelle:Letzte Seite URN DOI Abteilungen OPUS4-5156 Dissertation Castillo Alejandre, Susana Digital humanity: the temporal and semantic structure of dynamic conversational facial expressions This thesis focuses on establishing a - theoretically founded and empirically derived - novel methodological pipeline to provide Embodied Conversational Agents (ECAs) with natural facial expressions and desired personality traits. After giving an overview on the content of this thesis, we dedicate its second part to derive the Semantic Space (SSp) for facial expressions, finding that the same space is used for expression words, expression videos, and motion-capture-based point-cloud animations. The process involved the creation of a new facial expression database using Motion Capture (MoCap) technology. Our technique can be used to empirically map specific motion trajectories (including their frequency-decomposition) onto specific perceptual attributes, allowing the targeted creation of novel animations with the desired perceptual traits, as exemplified in the third part of this thesis. Before addressing our final conclusions and, on the grounds that the systematic differences between individuals while performing the same facial expressions are related to their personality, we devote the fourth part of this thesis to the study of the mapping between personality and expressive facial motions. 2019 urn:nbn:de:kobv:co1-opus4-51562 FG Praktische Informatik / Graphische Systeme OPUS4-462 Dissertation Kniemeyer, Ole Design and implementation of a graph grammar based language for functional-structural plant modelling Increasing biological knowledge requires more and more elaborate methods to translate the knowledge into executable model descriptions, and increasing computational power allows to actually execute these descriptions. Such a simulation helps to validate, extend and question the knowledge. For plant modelling, the well-established formal description language of Lindenmayer systems reaches its limits as a method to concisely represent current knowledge and to conveniently assist in current research. On one hand, it is well-suited to represent structural and geometric aspects of plant models - of which units is a plant composed, how are these connected, what is their location in 3D space -, but on the other hand, its usage to describe functional aspects - what internal processes take place in the plant structure, how does this interact with the structure - is not as convenient as desirable. This can be traced back to the underlying representation of structure as a linear chain of units, while the intrinsic nature of the structure is a tree or even a graph. Therefore, we propose to use graphs and graph grammars as a basis for plant modelling which combines structural and functional aspects. In the first part of this thesis, we develop the necessary theoretical framework. Starting with a presentation of the state of the art concerning Lindenmayer systems and graph grammars, we develop the formalism of relational growth grammars as a variant of graph grammars. We show that this formalism has a natural embedding of Lindenmayer systems which keeps all relevant properties, but represents branched structures directly as axial trees and not as linear chains with indirect encoding of branches. In the second part, we develop the main practical result, the XL programming language as an extension of the Java programming language by very general rule-based features. Short examples illustrate the application of the new language features. We describe the built-in pattern matching algorithm of the implemented run-time system for the XL programming language, and we sketch a possible implementation of an XL compiler. The third part is an application of relational growth grammars and the XL programming language. We show how the general XL interfaces can be customized for relational growth grammars. On top of this customization, several examples from a variety of disciplines demonstrate the usefulness of the developed formalism and language to describe plant growth, especially functional-structural plant models, but also artificial life, architecture or interactive games. Some examples operate on custom graphs like XML DOM trees or scene graphs of commercial 3D modellers, while the majority uses the 3D modelling platform GroIMP, a software developed in conjunction with this thesis. The appendix gives an overview of the GroIMP software. The practical usage of its plug-in for relational growth grammars is also illustrated. 2008 urn:nbn:de:kobv:co1-opus-5937 FG Praktische Informatik / Graphische Systeme OPUS4-5604 Dissertation Legde, Katharina Projecting motion capture : designing and implementing a modular and flexible facial animation pipeline to evaluate different perceptual effects We can not not communicate, humans use body language and facial expressions to communicate verbally and non-verbally with others. Humans are experts at deciphering and understanding facial expressions. Thus, synthesizing them on a virtual face becomes a very challenging task. Realistic facial animations have various applications like movies, games or affective interfaces. Often motion capture recordings of real humans are used to provide the virtual face with realistic motion. Animation controls need to be established to transfer the captured motion onto the virtual face. To mimic the full essence of the captured motion, these animation controls need to be carefully planned beforehand which is accompanied with high effort in terms of spent time and money. The presented thesis would like to offer an alternative approach to the common performance-driven facial animation techniques. Instead of using motion capture to drive a pre-defined blend-shape rig, this thesis projects motion capture directly onto a facial mesh. Throughout the proposed pipeline, different methods for retargeting, rigging and skinning are evaluated. Special attention is given to the fact that the origin of motion capture and the appearance of the virtual face does not need to be identical. A constraint for this approach to work is clean motion capture data. For that, this thesis offers an automatic way to clean facial motion trajectories and establishes stable and coherent motion curves. To not just give insights about the capability of the proposed pipeline but also to provide valuable results in the field of perception of virtual avatars, perceptual experiments are conducted. These experiments reveal the functionality of the pipeline and show that it is possible to re-use motion capture on different virtual faces. Additionally, the proposed pipeline is used as a tool to investigate into the perception of non-verbal communication for virtual avatars. 2021 urn:nbn:de:kobv:co1-opus4-56043 10.26127/BTUOpen-5604 FG Praktische Informatik / Graphische Systeme