Computer Graphics and Virtual Reality Lab

This paper addresses the problem of real-time joint localisation of legged skeletons in the presence of missing data. The data is assumed to be labelled 3d marker positions from a motion capture system. An integrated framework is presented which predicts the occluded marker positions using a variable turn model within an Unscented Kalman filter. Inferred information from neighbouring markers is used as observation states; these constraints are efficient, simple and real-time implementable. This work also takes advantage of the common case that missing markers are often visible to only a single camera, resulting in more accurate predictions. An Inverse Kinematics technique is then applied ensuring that the bone lengths remain constant over time; the system can thereby maintain a continuous data-flow. The marker and Centre of Rotation (CoR) positions can be calculated with high accuracy even in cases where markers are occluded for a long period of time. Our methodology is tested against some of the most popular methods for marker prediction and the results confirm that our approach outperforms these methods in estimating both marker and CoR positions.

Inverse Kinematics is defined as the problem of determining a set of appropriate joint configurations for which the end effectors move to desired positions as smoothly, rapidly, and as accurately as possible. However, many of the currently available methods suffer from high computational cost and production of unrealistic poses. In this paper, a novel heuristic method, called Forward And Backward Reaching Inverse Kinematics (FABRIK), is described and compared with some of the most popular existing methods regarding reliability, computational cost and conversion criteria. FABRIK avoids the use of rotational angles or matrices, and instead finds each joint position via locating a point on a line. Thus, it converges in fewer iterations, has low computational cost and produces visually realistic poses. Constraints can easily be incorporated within FABRIK and multiple chains with multiple end effectors are also easily supported.

This paper describes a novel iterative Inverse Kinematics (IK) solver, named FABRIK, that is implemented using Conformal Geometric Algebra (CGA). FABRIK uses a forward and backward iterative approach, finding each joint position via locating a point on a line. We also use the IK of a human hand as an example of implementation where a constrained version of FABRIK was employed for pose tracking. The hand is modelled using CGA, taking advantage of CGA's compact and geometrically intuitive framework and that basic entities in CGA, such as spheres, lines, planes and circles, are simply represented by algebraic objects. This approach can be used in a wide range of computer animation applications and is not limited to the specific problem discussed here. The proposed hand pose tracker is real-time implementable and exploits the advantages of CGA for applications in computer vision, graphics and robotics.

In this article an initiative to preserve the Cypriot folk dance heritage is reported. The project aims to create a publicly accessible digital archive of folk dances that does not only include video recordings, commonly used to document dance performances. In addition to rare video material held by local cultural institutions, state-of-the-art motion capture technologies are utilized to record and archive high quality motion data of expert dancers performing these traditional dances. Apart from the goal of preserving this intangible cultural heritage by digitizing it, the project is interested in increasing the awareness of the local community to its dance heritage. To achieve this a 3D video game for children is developed to teach these folk dances to the younger generations.

Articulated hand tracking systems have been commonly used in virtual reality applications, including systems with human-computer interaction or interaction with game consoles. However, building an effective real-time hand pose tracker remains challenging. In this , we present a simple and efficient methodology for tracking and reconstructing 3d hand poses using a markered optical motion capture system. Markers were positioned at strategic points, and an inverse kinematics solver was incorporated to fit the rest of the joints to the hand model. The model is highly constrained with rotational and orientational constraints, allowing motion only within a feasible set. The method is real-time implementable and the results are promising, even with a low frame rate.

This paper addresses the problem of real-time location of the joints or centres of rotation (CoR) of human skeletons in the presence of missing data. The data is assumed to be 3d marker positions from a motion capture system. We present an integrated framework which predicts the occluded marker positions using a Kalman filter in combination with inferred information from neighbouring markers and thereby maintains a continuous data-flow. The CoR positions can be calculated with high accuracy even in cases where markers are occluded for a long period of time.

This paper considers the problem of taking marker locations from optical motion capture data to identify and parameterise the underlying human skeleton structure and motion over time. It is concerned with real-time algorithms suitable for use within a visual feedback system. A common problem in motion capture is marker occlusion. Most current methods are only useful for offline processing or become ineffective when a significant portion of markers are missing for a long period of time. This paper presents a prediction algorithm, using a Kalman filter approach in combination with inferred information from neighbouring markers, to provide a continuous flow of data. The results are accurate and reliable even in cases where all markers on a limb are occluded, or one or two markers are not visible for a large sequence of frames. Pre-defined models are not required and skeleton fitting to this complete data can then be updated in real-time.

An architecture system and a method for tracking people are presented for sports applications. The system's input is video data from static camera and the output is the real world, real-time positions of the players during a sport event. This output can be used for low-bandwidth match play animations for web or wireless display, and for analysis of fitness and tactics of the teams and players. Firstly, an efficient real-time background modeling and maintenance is described based on the segmentation of input images into stationary and non-stationary blocks. In order to detect foreground objects a method based on background subtraction is implemented, using chromaticity and lightness, with the intention of avoiding shadows and jitters. Object tracking is achieved using a cost function of blob information such as the centroid coordinates, the covered area, the velocity and the color. A condensation filter is also implemented in order to cope with complex cases such as when regions enter/exit the scene and when they can be occluded by other regions. The results of the proposed methods, the players' location and trajectories are presented and discussed.

This work studies the problem of violence detection in audio data, which can be used for automated content rating. We employ some popular frame-level audio features both from the time and frequency domain. Afterwards, several statistics of the calculated feature sequences are fed as input to a Support Vector Machine classifier, which decides about the segment content with respect to violence. The presented experimental results verify the validity of the approach and exhibit a better performance than the other known approaches.

Η εργασία αυτή επικεντρώνεται στην μελέτη των μεθόδων που αποσκοπούν στην αυτόματη και αξιόπιστη αναγνώριση των σκηνών βίας με χρήση της ακουστικής πληροφορίας του σήματος. Χρησιμοποιήθηκαν συγκεκριμένα δημοφιλή ηχητικά χαρακτηριστικά όπως η ενεργειακή εντροπία, η ενέργεια μικρής διάρκειας (short time energy), το ZCR, η φασματική διακύμανση (spectral flux). Στη συνέχεια, αναπτύξαμε ένα αποτελεσματικό μηχανισμό ταξινόμησης SVM (Support Vector Machine), ο οποίος εκπαιδεύθηκε αξιοποιώντας αυτά τα ηχητικά χαρακτηριστικά γνωρίσματα της βίας. Τα αποτελέσματα αυτής της ταξινόμησης καταγράφονται, αξιολογούνται οι αλγόριθμοι και αναφέρονται η απόδοση και το ποσοστό επιτυχούς ταξινόμησης, το οποίο είναι καλύτερο από οποιαδήποτε άλλη προτεινόμενη μέθοδο.

Inverse Kinematics (IK) is defined as the problem of determining a set of appropriate joint configurations for which the end effectors move to desired positions as smoothly, rapidly, and as accurately as possible. During the last decades, several methods and techniques, sophisticated or heuristic, have been presented to produce fast and realistic solutions to the IK problem. However, most of the currently available methods suffer from high computational cost and production of unrealistic poses. This report reviews and compares the most popular IK methods regarding reliability, computational cost and conversion criteria, with a new heuristic iterative method, called Forward And Backward Reaching Inverse Kinematics (FABRIK). FABRIK avoids the use of rotational angles or matrices, and instead finds each joint position via locating a point on a line. Thus, it converges in fewer iterations, has low computational cost and produces realistic poses. Constraints can easily be incorporated within the FABRIK methodology and multiple chains with multiple end effectors are also easily supported.

This report considers the problem of using marker locations from optical motion capture data to identify and parameterise the underlying human skeleton structure and motion over time. It is concerned with real-time algorithms suitable for use within a visual feedback system. The algorithms presented require 3 markers on each limb segment and can be implemented in a sequential fashion; hence the computational cost of updating the centres of rotation is independent of the number of data points previously available. However, a common problem in motion capture is marker occlusion. Most current methods are only useful for offline processing or become ineffective when a significant proportion of markers are missing for a long period of time. This report presents an integrated framework which predicts the occluded marker positions using a Kalman filter in combination with inferred information from neighbouring markers and thereby maintains a continuous data-flow. The results are accurate and reliable even in cases where all markers on a limb segment are occluded, or one or two markers are non-visible for a large sequence of frames. Moreover, it investigates several extensions of the proposed methodology for real-time applications using a sophisticated variable turn model, together with rotor prediction and unscented theory for filtering. It compares each individual method according to its accuracy, complexity and processing time and suggests the optimal solution for each case. Pre-defined models are not required and skeleton fitting to this complete data can then be updated in real-time.

This project considers the problem of taking marker locations from optical motion capture data to identify and parameterise the underlying human skeleton structure and motion over time. It is concerned with real-time algorithms suitable for use within a visual feedback system. The algorithms presented require 3 markers on each limb segment and can be implemented in a sequential fashion; hence the computational cost of updating the centres of rotation is independent of the number of data points previously available. However, a common problem in motion capture is marker occlusion. Most current methods are only useful for offline processing or become ineffective when a significant proportion of markers are missing for a long period of time. This paper presents an integrated framework which predicts the occluded marker positions using a Kalman filter in combination with inferred information from neighbouring markers and thereby maintains a continues data-flow. The results are accurate and reliable even in cases where all markers on a limb segment are occluded, or one or two markers are non-visible for a large sequence of frames. Pre-defined models are not required and skeleton fitting to this complete data can then be updated in real-time.

An architecture system and a method for tracking people are presented for sports applications. The system's input is video data from static camera of a stadium and the output is the real world, real-time positions of the players during a sport event. This output can be used for low-bandwidth match play animations for web or wireless display, and for analysis of fitness and tactics of the teams and players. Firstly, in this study, a real-time background modelling and maintenance is described based on the segmentation of input images into stationary and non-stationary blocks. This method is efficient even if many objects are simultaneously visible and the background can be seen only for a short time of period. In order to detect foreground objects a method based on background subtraction is implemented, using chromaticity and lightness, with the intention of avoiding shadows and jitters. Furthermore, object tracking is achieved using a cost function of blob information such as the centroid coordinates, the covered area, the velocity and the colour. In addition to this, a condensation filter is implemented in order to cope with complex cases such as when regions enter/exit the scene and when they can be occluded by other regions. Thereafter, a field estimation algorithm, which utilizes the colour information of the ground is proposed and a new robust projective transformation from real image coordinates to the upper-view, virtual ground image is applied. Finally, the results of the proposed methods are presented individually in addition to the trajectories of moving objects in web/wireless applications such as a Java Applet.