Paul Newbury

Real Time Online Motion Capture for Entertainment Applications

We present a real-time online animation and lip syncing solution based on personal motion sensing for use in entertainment and theme park applications. Our solution streams real-time motion capture based animation, video and voice data over the Internet using UDP. Motion capture data is streamed from the eMove mocap suit over the Internet into a client application, which provides a lip synced 3D animated character or puppet. Applications include online digital puppeteering and games where the character is animated in real time.

Why is my home not smart

Although the idea of the smart home has been around for over three decades the smart technology that enables it has yet to reach the mass market. Spending on smart technology is expected to rise, but it is still negligible when compared to overall spending on consumer electronics. This paper examines the benefits of the smart home, peoples attitude towards them and smart technologies and the possible reasons for lack of interest and adoption of such technologies.

Evaluating the covariance matrix constraints for data-driven statistical human motion reconstruction

This paper presents the evaluation process of the characters motion reconstruction while constraints are applied to the covariance matrix of the motion prior learning process. For the evaluation process, a maximum a posteriori (MAP) framework is first generated, which receives input trajectories and reconstructs the motion of the character. Then, using various methods to constrain the covariance matrix, information that reflects certain assumptions about the motion reconstruction process is retrieved. Each of the covariance matrix constraints are evaluated by its ability to reconstruct the desired motion sequences either by using a large amount of motion data or by using a small dataset that contains only specific motions.

An open source Grid based render farm for Blender 3D

Growing demand for scientific simulations that visualise results through 3D graphics requires high quality 3D models and animations. Better quality 3D models, e.g. more complex geometry and photo-realistic rendering, always need more time and computational power for rendering. The rendering process assumes time and power are inversely proportional to each other, the less computational power you have the more time you require to render and the more power you have the less time you need to render. Traditionally, a render farm requires a cluster of computers to perform satisfactory rendering in terms of computational power and time, but not every scientific organization can afford such clusters. This paper presents an open source Grid based render farm, which can be plugged into Blender 3D as a tool bar that exploits a Condor based Grid.

How to Build an Open Source Render Farm based on Desktop Grid Computing

This paper presents an experiment on how to implement a Grid-based High Performance Computing solution using existing resources typically available in a teaching or research laboratory. A cost-effective solution is proposed based on open source software components, and, where appropriate, our own software solutions, for large scientific applications in the public sector such as universities and research institutes. In such institutions, classical solutions for HPC are often not affordable, yet they usually have at their disposal a large number of machines that can be utilised. The Department of Informatics at University of Sussex, for example, has just installed 150 new Core2 Duo machines across 3 laboratories. By scaling this number up across the whole University, it can result a large potential computing resource for utilization. Typical processor usage rates are often somewhere between 10% and 20% (i.e. user-generated processes) for most machines. This paper proposes a solution that exploits the remaining 80% to 90% processor power through consumption of available computer idle time without disturbing current users. To achieve this goal, the open source Condor High Throughput Computing software was selected and implemented as a desktop Grid computing solution. This paper presents our experiences in finding a solution so that other institutions can develop similar Grid solutions for their own large scientific experiments, taking advantage of their existing resources. The implementation of our solution is analyzed in the context of building a render farm.

Data-Driven Motion Reconstruction Using Local Regression Models

Reconstructing human motion data using a few input signals or trajectories is always challenging problem. This is due to the difficulty of reconstructing natural human motion since the low-dimensional control parameters cannot be directly used to reconstruct the high-dimensional human motion. Because of this limitation, a novel methodology is introduced in this paper that takes benefit of local dimensionality reduction techniques to reconstruct accurate and natural-looking full-body motion sequences using fewer number of input. In the proposed methodology, a group of local dynamic regression models is formed from pre-captured motion data to support the prior learning process that reconstructs the full-body motion of the character. The evaluation that held out has shown that such a methodology can reconstruct more accurate motion sequences than possible with other statistical models.

Splicing of Concurrent Upper-body Motion Spaces with Locomotion

In this paper, we present a motion splicing technique for generating concurrent upper-body actions occurring simultaneously with the evolution of a lower-body locomotion sequence. Specifically, we show that a layered interpolation motion model generates upper-body poses while assigning different actions to each upper-body part. Hence, in the proposed motion splicing approach, it is possible to increase the number of generated motions as well as the number of desired actions that can be performed by virtual characters. Additionally, we propose an iterative motion blending solution, inverse pseudo-blending, to maintain a smooth and natural interaction between the virtual character and the virtual environment; inverse pseudo-blending is a constraint-based motion editing technique that blends the motions enclosed in a tetrahedron by minimising the distances between the end-effector positions of the actual and blended motions. Additionally, to evaluate the proposed solution, we implemented an example-based application for interactive motion splicing based on specified constraints. Finally, the generated results show that the proposed solution can be beneficially applied to interactive applications where concurrent actions of the upper-body are desired.

LUT‐based compensation model for OLED degradation

A compensation model for blue OLED devices, which is based on an OLED degradation model derived from luminance measurements of OLED devices, has been developed; it reduces degradation effects by monitoring the activity of a set of subpixels and adjusting the driving conditions of the blue subpixels of the display accordingly. To evaluate its performance, the compensation model has been embedded in an OLED display controller model, where its implementation is based on a lookup table. Simulations show that even with a reduced number of monitored subpixels, degradation is effectively reduced. B)2005 Society for Information Display

Electronic simulation for virtual reality: virtual prototyping

This paper examines a virtual prototyping system for electronic devices which incorporate visualisation using a novel integrated development environment that combines user interaction with photorealistic 2D and 3D models. Full system level hardware simulation is also supported within this framework which offers electronic simulation in a virtual environment. This helps to link product development specialists with a unified and coherent modelling environment. Virtual prototyping is a novel design methodology that aims to decrease the time-to-market and increase product reliability, quality and fulfilment of user requirements. This paper uses the example of a remotely controlled domestic cooking system to illustrate this process

Finger motion estimation and synthesis for gesturing characters

In this paper, a novel methodology for estimating a virtual characters finger motion based on parameters provided by the characters wrist is presented. For the motion estimation process, firstly the motion sequences are classified into active and passive finger gesture phases with the active finger gestures classified according to different gesture types. Based on both classifications the system first searches for the gesture phase and then for the most appropriate gesture type. Having found the gesture type, or having determined that the input motion segment belongs to the passive phase, by using a metric, it retrieves the closest motion segment. Such a method can be beneficial in the finger motion estimation process, since both wrong estimations and the computational time of the searching process are reduced. Finally, in addition to the motion estimation process, an optimisation of the motion graphs methodology for searching optimal transitions between two consecutive motions is introduced.