Automating the conversion of natural language fiction to multi-modal 3D animated virtual environments
- Authors: Glass, Kevin Robert
- Date: 2009
- Subjects: Virtual computer systems , Virtual storage (Computer science) , Virtual reality , Computer animation , Fiction -- Computer programs , Narration (Rhetoric) -- Computer simulation , Animation (Cinematography) , Natural language processing (Computer Science)
- Language: English
- Type: Thesis , Doctoral , PhD
- Identifier: vital:4632 , http://hdl.handle.net/10962/d1006518
- Description: Popular fiction books describe rich visual environments that contain characters, objects, and behaviour. This research develops automated processes for converting text sourced from fiction books into animated virtual environments and multi-modal films. This involves the analysis of unrestricted natural language fiction to identify appropriate visual descriptions, and the interpretation of the identified descriptions for constructing animated 3D virtual environments. The goal of the text analysis stage is the creation of annotated fiction text, which identifies visual descriptions in a structured manner. A hierarchical rule-based learning system is created that induces patterns from example annotations provided by a human, and uses these for the creation of additional annotations. Patterns are expressed as tree structures that abstract the input text on different levels according to structural (token, sentence) and syntactic (parts-of-speech, syntactic function) categories. Patterns are generalized using pair-wise merging, where dissimilar sub-trees are replaced with wild-cards. The result is a small set of generalized patterns that are able to create correct annotations. A set of generalized patterns represents a model of an annotator's mental process regarding a particular annotation category. Annotated text is interpreted automatically for constructing detailed scene descriptions. This includes identifying which scenes to visualize, and identifying the contents and behaviour in each scene. Entity behaviour in a 3D virtual environment is formulated using time-based constraints that are automatically derived from annotations. Constraints are expressed as non-linear symbolic functions that restrict the trajectories of a pair of entities over a continuous interval of time. Solutions to these constraints specify precise behaviour. We create an innovative quantified constraint optimizer for locating sound solutions, which uses interval arithmetic for treating time and space as contiguous quantities. This optimization method uses a technique of constraint relaxation and tightening that allows solution approximations to be located where constraint systems are inconsistent (an ability not previously explored in interval-based quantified constraint solving). 3D virtual environments are populated by automatically selecting geometric models or procedural geometry-creation methods from a library. 3D models are animated according to trajectories derived from constraint solutions. The final animated film is sequenced using a range of modalities including animated 3D graphics, textual subtitles, audio narrations, and foleys. Hierarchical rule-based learning is evaluated over a range of annotation categories. Models are induced for different categories of annotation without modifying the core learning algorithms, and these models are shown to be applicable to different types of books. Models are induced automatically with accuracies ranging between 51.4% and 90.4%, depending on the category. We show that models are refined if further examples are provided, and this supports a boot-strapping process for training the learning mechanism. The task of interpreting annotated fiction text and populating 3D virtual environments is successfully automated using our described techniques. Detailed scene descriptions are created accurately, where between 83% and 96% of the automatically generated descriptions require no manual modification (depending on the type of description). The interval-based quantified constraint optimizer fully automates the behaviour specification process. Sample animated multi-modal 3D films are created using extracts from fiction books that are unrestricted in terms of complexity or subject matter (unlike existing text-to-graphics systems). These examples demonstrate that: behaviour is visualized that corresponds to the descriptions in the original text; appropriate geometry is selected (or created) for visualizing entities in each scene; sequences of scenes are created for a film-like presentation of the story; and that multiple modalities are combined to create a coherent multi-modal representation of the fiction text. This research demonstrates that visual descriptions in fiction text can be automatically identified, and that these descriptions can be converted into corresponding animated virtual environments. Unlike existing text-to-graphics systems, we describe techniques that function over unrestricted natural language text and perform the conversion process without the need for manually constructed repositories of world knowledge. This enables the rapid production of animated 3D virtual environments, allowing the human designer to focus on creative aspects.
- Full Text:
- Date Issued: 2009
Development of the components of a low cost, distributed facial virtual conferencing system
- Authors: Panagou, Soterios
- Date: 2000 , 2011-11-10
- Subjects: Virtual computer systems , Virtual reality , Computer conferencing
- Language: English
- Type: Thesis , Masters , MSc
- Identifier: vital:4622 , http://hdl.handle.net/10962/d1006490 , Virtual computer systems , Virtual reality , Computer conferencing
- Description: This thesis investigates the development of a low cost, component based facial virtual conferencing system. The design is decomposed into an encoding phase and a decoding phase, which communicate with each other via a network connection. The encoding phase is composed of three components: model acquisition (which handles avatar generation), pose estimation and expression analysis. Audio is not considered part of the encoding and decoding process, and as such is not evaluated. The model acquisition component is implemented using a visual hull reconstruction algorithm that is able to reconstruct real-world objects using only sets of images of the object as input. The object to be reconstructed is assumed to lie in a bounding volume of voxels. The reconstruction process involves the following stages: - Space carving for basic shape extraction; - Isosurface extraction to remove voxels not part of the surface of the reconstruction; - Mesh connection to generate a closed, connected polyhedral mesh; - Texture generation. Texturing is achieved by Gouraud shading the reconstruction with a vertex colour map; - Mesh decimation to simplify the object. The original algorithm has complexity O(n), but suffers from an inability to reconstruct concave surfaces that do not form part of the visual hull of the object. A novel extension to this algorithm based on Normalised Cross Correlation (NCC) is proposed to overcome this problem. An extension to speed up traditional NCC evaluations is proposed which reduces the NCC search space from a 2D search problem down to a single evaluation. Pose estimation and expression analysis are performed by tracking six fiducial points on the face of a subject. A tracking algorithm is developed that uses Normalised Cross Correlation to facilitate robust tracking that is invariant to changing lighting conditions, rotations and scaling. Pose estimation involves the recovery of the head position and orientation through the tracking of the triangle formed by the subject's eyebrows and nose tip. A rule-based evaluation of points that are tracked around the subject's mouth forms the basis of the expression analysis. A user assisted feedback loop and caching mechanism is used to overcome tracking errors due to fast motion or occlusions. The NCC tracker is shown to achieve a tracking performance of 10 fps when tracking the six fiducial points. The decoding phase is divided into 3 tasks, namely: avatar movement, expression generation and expression management. Avatar movement is implemented using the base VR system. Expression generation is facilitated using a Vertex Interpolation Deformation method. A weighting system is proposed for expression management. Its function is to gradually transform from one expression to the next. The use of the vertex interpolation method allows real-time deformations of the avatar representation, achieving 16 fps when applied to a model consisting of 7500 vertices. An Expression Parameter Lookup Table (EPLT) facilitates an independent mapping between the two phases. It defines a list of generic expressions that are known to the system and associates an Expression ID with each one. For each generic expression, it relates the expression analysis rules for any subject with the expression generation parameters for any avatar model. The result is that facial expression replication between any subject and avatar combination can be performed by transferring only the Expression ID from the encoder application to the decoder application. The ideas developed in the thesis are demonstrated in an implementation using the CoRgi Virtual Reality system. It is shown that the virtual-conferencing application based on this design requires only a bandwidth of 2 Kbps. , Adobe Acrobat Pro 9.4.6 , Adobe Acrobat 9.46 Paper Capture Plug-in
- Full Text:
- Date Issued: 2000