Learning Two-Person Interaction Models for Responsive Synthetic Humanoids

Imitation learning is a promising approach for generating life-like behaviors of virtual humans and humanoid robots. So far, however, imitation learning has been mostly restricted to single agent settings where observed motions are adapted to new environment conditions but not to the dynamic behavior of interaction partners. In this paper, we introduce a new imitation learning approach that is based on the simultaneous motion capture of two human interaction partners. From the observed interactions, low-dimensional motion models are extracted and a mapping between these motion models is learned. This interaction model allows the real-time generation of agent behaviors that are responsive to the body movements of an interaction partner. The interaction model can be applied both to the animation of virtual characters as well as to the behavior generation for humanoid robots.

Hands-Free Navigation in Immersive Environments for the Evaluation of the Effectiveness of Indoor Navigation Systems

While navigation systems for cars are in widespread use, only recently, indoor navigation systems based on smartphone apps became technically feasible. Hence tools in order to plan and evaluate particular designs of information provision are needed. Since tests in real infrastructures are costly and environmental conditions cannot be held constant, one must resort to virtual infrastructures. This paper presents the development of an environment for the support of the design of indoor navigation systems whose center piece consists in a hands-free navigation method using the Microsoft Kinect in the four-sided Definitely Affordable Virtual Environment (DAVE). Navigation controls using the user's gestures and postures as the input to the controls are designed and implemented. The installation of expensive and bulky hardware like treadmills is avoided while still giving the user a good impression of the distance she has traveled in virtual space. An advantage in comparison to approaches using a head mounted display is that the DAVE allows the users to interact with their smartphone. Thus the effects of different indoor navigation systems can be evaluated already in the planning phase using the resulting system

A user supported object tracking framework for interactive video production

We present a user supported tracking framework that combines automatic tracking with extended user input to create error free tracking results that are suitable for interactive video production. The goal of our approach is to keep the necessary user input as small as possible. In our framework, the user can select between different tracking algorithms - existing ones and new ones that are described in this paper. Furthermore, the user can automatically fuse the results of different tracking algorithms with our robust fusion approach. The tracked object can be marked in more than one frame, which can significantly improve the tracking result. After tracking, the user can validate the results in an easy way, thanks to the support of a powerful interpolation technique. The tracking results are iteratively improved until the complete track has been found. After the iterative editing process the tracking result of each object is stored in an interactive video file that can be loaded by our player for interactive videos.

Collision Detection: Broad Phase Adaptation from Multi-Core to Multi-GPU Architecture

We present in this paper several contributions on the collision detection optimization centered on hardware performance. We focus on the broad phase which is the first step of the collision detection process and propose three new ways of parallelization of the well-known Sweep and Prune algorithm. We first developed a multi-core model takes into account the number of available cores. Multi-core architecture enables us to distribute geometric computations with use of multi-threading. Critical writing section and threads idling have been minimized by introducing new data structures for each thread. Programming with directives, like OpenMP, appears to be a good compromise for code portability. We then proposed a new GPU-based algorithm also based on the "Sweep and Prune" that has been adapted to multi-GPU architectures. Our technique is based on a spatial subdivision method used to distribute computations among GPUs. Results show that significant speed-up can be obtained by passing from 1 to 4 GPUs in a large-scale environment.

Comparison of 2D and 3D GUI Widgets for Stereoscopic Multitouch Setups

Recent developments in the area of interactive entertainment have suggested to combine stereoscopic visualization with multi-touch displays, which has the potential to open up new vistas for natural interaction with interactive three-dimensional (3D) applications. However, the question arises how the user interfaces for system control in such 3D setups should be designed in order to provide an effective user experience. In this article we introduce 3D GUI widgets for interaction with stereoscopic touch displays. The design of the widgets was inspired to skeuomorphism and affordances in such a way that the user should be able to operate the virtual objects in the same way as their real-world equivalents. We evaluate the developed widgets and compared them with their 2D counterparts in the scope of an example application in order to analyze the usability of and user behavior with the widgets. The results reveal differences in user behavior with and without stereoscopic display during touch interaction, and show that the developed 2D as well as 3D GUI widgets can be used effectively in different applications.

Estimating Gesture Accuracy in Motion-Based Health Games

This manuscript details a technique for estimating gesture accuracy within the context of motion-based health video games using the MICROSOFT KINECT. We created a physical therapy game that requires players to imitate clinically significant reference gestures. Player performance is represented by the degree of similarity between the performed and reference gestures and is quantified by collecting the Euler angles of the player's gestures, converting them to a three-dimensional vector, and comparing the magnitude between the vectors. Lower difference values represent greater gestural correspondence and therefore greater player performance. A group of thirty-one subjects was tested. Subjects achieved gestural correspondence sufficient to complete the game's objectives while also improving their ability to perform reference gestures accurately.

A Comparative Evaluation of Three Skin Color Detection Approaches

Skin segmentation is a challenging task due to several influences such as unknown lighting conditions, skin colored background, and camera limitations. A lot of skin segmentation approaches were proposed in the past including adaptive (in the sense of updating the skin color online) and non-adaptive approaches. In this paper, we compare three skin segmentation approaches that are promising to work well for hand tracking, which is our main motivation for this work. Hand tracking can widely be used in VR/AR e.g. navigation and object manipulation. The first skin segmentation approach is a well-known non-adaptive approach. It is based on a simple, pre-computed skin color distribution. Methods two and three adaptively estimate the skin color in each frame utilizing clustering algorithms. The second approach uses a hierarchical clustering for a simultaneous image and color space segmentation, while the third approach is a pure color space clustering, but with a more sophisticated clustering approach. For evaluation, we compared the segmentation results of the approaches against a ground truth dataset. To obtain the ground truth dataset, we labeled about 500 images captured under various conditions.

Influence of Comfort on 3D Selection Task Performance in Immersive Desktop Setups

Immersive virtual environments (IVEs) have the potential to afford natural interaction in the three-dimensional (3D) space around a user. However, interaction performance in 3D mid-air is often reduced and depends on a variety of ergonomics factors, the user's endurance, muscular strength, as well as fitness. In particular, in contrast to traditional desktop-based setups, users often cannot rest their arms in a comfortable pose during the interaction. In this article we analyze the impact of comfort on 3D selection tasks in an immersive desktop setup. First, in a pre-study we identified how comfortable or uncomfortable specific interaction positions and poses are for users who are standing upright. Then, we investigated differences in 3D selection task performance when users interact with their hands in a comfortable or uncomfortable body pose, while sitting on a chair in front of a table while the VE was displayed on a headmounted display (HMD). We conducted a Fitts' Law experiment to evaluate selection performance in different poses. The results suggest that users achieve a significantly higher performance in a comfortable pose when they rest their elbow on the table.

Advanced luminance control and black offset correction for multi-projector display systems

In order to display a homogeneous image using multiple projectors, differences in the projected intensities must be compensated. In this paper, we present novel approaches to combine and extend existing techniques for edge blending and luminance harmonization to achieve a detailed luminance control. Furthermore, we apply techniques for improving the contrast ratio of multi-segmented displays also to the black offset correction. We also present a simple scheme to involve the displayed context in the correction process to dynamically improve the contrast in brighter images. In addition, we present a metric to evaluate the different methods and their influence on the visual quality.

Real-time depth camera tracking with CAD models and ICP

In recent years, depth cameras have been widely utilized in camera tracking for augmented and mixed reality. Many of the studies focus on the methods that generate the reference model simultaneously with the tracking and allow operation in unprepared environments. However, methods that rely on predefined CAD models have their advantages. In such methods, the measurement errors are not accumulated to the model, they are tolerant to inaccurate initialization, and the tracking is always performed directly in reference model's coordinate system. In this paper, we present a method for tracking a depth camera with existing CAD models and the Iterative Closest Point (ICP) algorithm. In our approach, we render the CAD model using the latest pose estimate and construct a point cloud from the corresponding depth map. We construct another point cloud from currently captured depth frame, and find the incremental change in the camera pose by aligning the point clouds. We utilize a GPGPU-based implementation of the ICP which efficiently uses all the depth data in the process. The method runs in real-time, it is robust for outliers, and it does not require any preprocessing of the CAD models. We evaluated the approach using the Kinect depth sensor, and compared the results to a 2D edge-based method, to a depth-based SLAM method, and to the ground truth. The results show that the approach is more stable compared to the edge-based method and it suffers less from drift compared to the depth-based SLAM.