Figure 1: Examples from our test set of motions. The left two images are natural (motion capture data). The two images to the right are unnatural (badly edited and incompletely cleaned motion). Joints that are marked in red-yellow were detected as having unnatural motion. Frames for these images were selected by the method presented in [Assa et al. 2005]. In this paper, we investigate whether it is possible to develop a measure that quantifies the naturalness of human motion (as defined by a large database). Such a measure might prove useful in verifying that a motion editing operation had not destroyed the naturalness of a motion capture clip or that a synthetic motion transition was within the space of those seen in natural human motion. We explore the performance of mixture of Gaussians (MoG), hidden Markov models (HMM), and switching linear dynamic systems (SLDS) on this problem. We use each of these statistical models alone and as part of an ensemble of smaller statistical models. We also implement a Naive Bayes (NB) model for a baseline comparison. We test these techniques on motion capture data held out from a database, keyframed motions, edited motions, motions with noise added, and synthetic motion transitions. We present the results as receiver operating characteristic (ROC) curves and compare the results to the judgments made by subjects in a user study.

Animation techniques for controlling passive simulation are commonly based on an optimization paradigm: the user provides goals a priori, and sophisticated numerical methods minimize a cost function that represents these goals. Unfortunately, for multibody systems with discontinuous contact events these optimization problems can be highly nontrivial to solve, and many-hour offline optimizations, unintuitive parameters, and convergence failures can frustrate end-users and limit usage. On the other hand, users are quite adaptable, and systems which provide interactive feedback via an intuitive interface can leverage the user’s own abilities to quickly produce interesting animations. However, the online computation necessary for interactivity limits scene complexity in practice. We introduce Many-Worlds Browsing, a method which circumvents these limits by exploiting the speed of multibody simulators to compute numerous example simulations in parallel (offline and online), and allow the user to browse and modify them interactively. We demonstrate intuitive interfaces through which the user can select among the examples and interactively adjust those parts of the scene that do not match his requirements. We show that using a combination of our techniques, unusual and interesting results can be generated for moderately sized scenes with under an hour of user time. Scalability is demonstrated by sampling much larger scenes using modest offline computations.

In large-scale simulations involving complex scenes, such as cities inhabited by crowds, simplifications are almost always necessary to achieve interactive frame-rates. Level of Detail (LOD) techniques such as reducing geometric complexity, or substituting impostor images for geometry, are usually employed. Image-based or impostor techniques have been gaining in popularity in recent years, along with hybrid methods that combine impostors and geometry, but perceptual issues with respect to such representations have been largely neglected to date. In this paper we evaluate the effectiveness of impostor representations for the real-time rendering and animation of static buildings and dynamic virtual humans. Using sets of psychophysical experiments, we establish some thresholds at which impostors are effective for static and dynamic objects, along with criteria for selecting transitions to geometry and update rates. We also compare the impact of two model representations (geometry and impostor), on the perception of human motion. We have found that impostors are an extremely effective substitute for detailed geometry in the target application area. Categories and Subject Descriptors (according to ACM CCS): I.3.7 [Computer Graphics]: Virtual Reality, Perception 1.

In multimedia applications, it is essential to distribute resources efficiently among different types of data in order to optimize overall quality. We propose a perceptual metric using Just-Noticeable-Difference (JND) to identify redundant mesh data so that available bandwidth can be allocated to improve texture resolution. Evaluation of perceptual impact during runtime is based on statistics in a lookup table generated during preprocessing. If the impact is less than the JND, no mesh refinement is performed. We apply Weber’s fraction to compute the JND threshold, which is verified by perceptual evaluations. Experimental result shows that our JND model can accurately predict perceptual impact based on the human visual system. 1. Introduction structures tend to diminish as σ increases, and only major structures survive at higher scales (Fig. 1). When transmitting 3D textured mesh (TexMesh) over a Decimation and refinement are performed using edge shared network, limited resources such as bandwidth has to collapse and vertex split operations. A detailed discussion of be allocated between both mesh and texture data. various mesh simplification approaches can be found in

Most of current facial animation approaches largely focus on the accuracy or efficiency of their algorithms, or how to optimally utilize pre-collected facial motion data. However, human perception, the ultimate measuring stick of the visual fidelity of synthetic facial animations, was not effectively exploited in these approaches. In this paper, we present a novel perceptually guided computational framework for expressive facial animation, by bridging objective facial motion patterns with subjective perceptual outcomes. First, we construct a facial perceptual metric (FacePEM) using a hybrid of region-based facial motion analysis and statistical learning techniques. The constructed FacePEM model can automatically measure the emotional expressiveness of a facial motion sequence. We showed how the constructed FacePEM model can be effectively incorporated into various facial animation algorithms. For the sake of clear demonstrations, we choose data-driven expressive speech animation generation and expressive facial motion editing as two concrete application examples. Through a comparative user study, we showed that comparing with the traditional facial animation algorithms, the introduced perceptually guided expressive facial animation algorithms can significantly increase the emotional expressiveness and perceptual believability of synthesized facial animations.

Complex acrobatic stunts, such as double or triple flips, can be performed only by highly skilled athletes. On the other hand, simpler tricks, such as single-flip jumps, are relatively easy to master. We present a method for creating complex, multi-flip ballistic motions from simple, single-flip jumps. Our approach also allows an animator to interact with the system by introducing modifications to a ballistic phase of a motion. Our method automatically adjusts motion trajectories, to assure physical validity of the motion after the modifications. The presented technique is efficient and produces physically valid results without resorting to computationally expensive optimization. To validate our approach we present the results of a study of user sensitivity to errors in angular momentum and take-off angle. The study shows that small changes of these parameters introduced by our method are not perceptible to a viewer.

A deeper understanding of what makes animation perceptually plausible would benefit a number of applications, such as approximate collision detection and goal-directed animation. In a series of psychophysical experiments, we examine how measurements of perceptual sensitivity in realistic physical simulations compare to similar measurements done in more abstract settings. We find that participant tolerance for certain types of errors is significantly higher in a realistic snooker scenario than in the abstract test settings previously used to examine those errors. By contrast, we find tolerance for errors displayed in realistic but more neutral environments was not different from tolerance for those errors in abstract settings. Additionally, we examine the interaction of auditory and visual cues in determining participant sensitivity to spatiotemporal errors in rigid body collisions. We find that participants are predominantly affected by visual cues. Finally, we find that tolerance for spatial gaps during collision events is constant for a wide range of viewing angles if the effect of foreshortening and occlusion caused by the viewing angle is taken into account.

To my wife Wei and my son Billy. iv Generating human motion that appears natural is a long standing problem in character animation. Researchers have explored many different approaches including physics-based simulation, optimization, and data-driven methods such as motion graphs and motion interpolation. One major difficulty in applying most of these approaches is the lack of an implementable definition of what it means for motion to be natural or human-like. In this thesis, we explore two techniques to fill this gap. The first technique creates a naturalness measure for quantifying natural human motion. The second technique involves a statistical analysis of human motion to compute aggregate statistics that are needed to guide animation algorithms for human figures toward natural looking solutions. A naturalness measure should be useful in verifying that a motion editing

An increasing number of projects have examined the perceptual magnitude of visible artifacts in animated motion. These studies have been performed using a mix of character types, from detailed human models to abstract geometric objects such as spheres. We explore the extent to which character morphology influences user sensitivity to errors in a fixed set of ballistic motions replicated on three different character types. We find user sensitivity responds to changes in error type or magnitude in a similar manner regardless of character type, but that users display a higher sensitivity to some types of errors when these errors are displayed on more human-like characters. Further investigation of those error types suggests that being able to observe a period of preparatory motion before the onset of ballistic motion may be important. However, we found no evidence to suggest that a mismatch between the preparatory phase and the resulting ballistic motion was responsible for the higher sensitivity to errors that was observed for the most humanlike character. Categories and Subject Descriptors (according to ACM CCS):

A non-isotropic hybrid image/model based gaze-contingent rendering technique utilizing ray casting on a GPU is discussed. Empirical evidence derived from human subject experiments indicates an inverse relationship between a peripherally degraded scene’s high-resolution inset size and mean search time, a trend consis-tent with existing image-based and model-based techniques. In addition, the data suggest that maintaining a target’s silhouette edges decreases search times when compared to targets with degraded edges. Benefits of the hybrid technique include simplicity of design and parallelizability, both conducive to GPU implementation.