Fast stochastic optimization for articulated structure tracking
Image and Vision Computing
Learning-based dynamic coupling of discrete and continuous trackers
Computer Vision and Image Understanding - Special issue on modeling people: Vision-based understanding of a person's shape, appearance, movement, and behaviour
Vision-based hand pose estimation: A review
Computer Vision and Image Understanding
Nearest neighbor search methods for handshape recognition
Proceedings of the 1st international conference on PErvasive Technologies Related to Assistive Environments
A Single Camera Motion Capture System for Human-Computer Interaction
IEICE - Transactions on Information and Systems
A robust and accurate 3d hand posture estimation method for interactive systems
Proceedings of the fourth international conference on Tangible, embedded, and embodied interaction
A variational approach to monocular hand-pose estimation
Computer Vision and Image Understanding
Hand posture estimation in complex backgrounds by considering mis-match of model
ACCV'07 Proceedings of the 8th Asian conference on Computer vision - Volume Part I
A database-based framework for gesture recognition
Personal and Ubiquitous Computing
Silhouette area based similarity measure for template matching in constant time
AMDO'10 Proceedings of the 6th international conference on Articulated motion and deformable objects
3-D hand posture recognition by training contour variation
FGR' 04 Proceedings of the Sixth IEEE international conference on Automatic face and gesture recognition
Segmentation-free, area-based articulated object tracking
ISVC'11 Proceedings of the 7th international conference on Advances in visual computing - Volume Part I
3D hand pose reconstruction with ISOSOM
ISVC'05 Proceedings of the First international conference on Advances in Visual Computing
Template-Based hand pose recognition using multiple cues
ACCV'06 Proceedings of the 7th Asian conference on Computer Vision - Volume Part II
Real-time viewpoint-invariant hand localization with cluttered backgrounds
Image and Vision Computing
Template-based hand detection and tracking
ASB'03 Proceedings of the 1st international conference on Advanced Studies in Biometrics
Real-Time 3d hand shape estimation based on inverse kinematics and physical constraints
ICIAP'05 Proceedings of the 13th international conference on Image Analysis and Processing
Dynamic data driven coupling of continuous and discrete methods for 3d tracking
ICCS'05 Proceedings of the 5th international conference on Computational Science - Volume Part II
An image based algorithm to safely locate human extremities for human-robot collaboration
ICIRA'12 Proceedings of the 5th international conference on Intelligent Robotics and Applications - Volume Part II
| Hi-index | 0.00 |
Abstract: This paper proposes a system estimating arbitrary 3-D human hand postures in real-time. Differently from most of the previous real-time system, it can accept not only pre-determined hand signs but also arbitrary postures and it works in a monocular camera environment. The estimation is based on a 2-D image retrieval. More than 16000 possible hand appearances (silhouette contours) are first generated from a given 3-D shape model by rotating model joints and stored in an appearance database. Every appearance is tagged with its own joint angles which are used when the appearance was generated. By retrieving the appearance in the database well-matching to the input image contour, the joint angles of the input shape can be rapidly obtained. For robust matching, little difference between the appearances and the input is adjusted before matching considering shape deformation and quantization error in appearence samping. In order to achieve the real-time processing, the search area is reduced by using an adjacency map in the database. In the map, adjacent appearances having similar joint angles are connected with each other. For each appearance, a neighborhood which consists of the adjacent appearances is defined. In each frame, the search area is limited to the neighborhood of the estimated appearance in the previous frame. The best candidate at one frame is sometimes wrong due to approximation errors, too rapid motions or ambiguities caused by self-occlusion. To prevent tracking failures, a fixed number of the well-matching appearances are saved at every frame. After the multiple neighborhoods of the saved appearances are merged, the unified neighborhood is searched for the estimate efficiently by Beam search. These algorithms are implemented on a PC cluster system consisting of 6 PCs (Pentium III 600MHz) and real-time processing is achieved. The resulted posture estimates are shown by experimental examples.