Holden E., "Visual Recognition of Hand Motion", PhD Thesis, Department of Computer Science, University of Western Australia, January 1997  Home/Search   Document Details and Download   Summary   Related Articles   Check

....The model used depends on the application. In case of gesture recognition [35] which is a more extensively studied research area, one approach is representing the hand pose using view dependent appearance based (2D) models [12] another approach is using view independent 3D hand models [37] [24], 16] For manipulation type of applications, which is the main characteristics of this project, the use of a 3D hand model is a requirement. Modeling the hand especially in a real time application is not a very easy problem to be solved because of the complexity of the human hand structure [2] ....

....Only approximate solutions can be found. 16 . Mapping from 2D Data to 3D Angles [39] 10] In this approach the joint angles and locations on the 2D image data or the whole image itself are used to drive a mapping a mapping from the 2D data to 3D joint angles. Top down Approach [16] [24], 37] In the top down approach the 3D model is projected onto the image(s) and the projection is compared with the image data. An optimization procedure where the residual between the features in the image and the projected model is minimized is executed. The model parameters yielding the ....

[Article contains additional citation context not shown here]

Holden E., "Visual Recognition of Hand Motion", PhD Thesis, Department of Computer Science, University of Western Australia, January 1997

....yet al..so supports data classification over a range of errors in the motion sensing process or slight individual hand movement variations. This is achieved by using an adaptive fuzzy expert system. Since then, we have applied this technique to a vision based Hand Motion Understanding (HMU) system [15] in order to recognise static and dynamic Australian Sign Language (Auslan) signs that use fine grain hand movement such as movement of the fingers. Unlike other visionbased systems, the HMU system performs 3D model based tracking on the image sequence from a single viewpoint, producing the ....

E. J. Holden,Visual Recognition of Hand Motion, PhD thesis, University of Western Australia, 1997.

....to overcome this limitation, Watanabe and Yachida [10] approximate 3D information by using an eigenspace constructed from multiple input sequences that are captured from many directions, without reconstructing 3D structure. We have developed the vision based Hand Motion Understanding (HMU) system [11] that extracts and classifies 3D hand configuration data from images taken from a single viewpoint, in order to understand static and dynamic Australian Sign Language (Auslan) signs. Throughout this paper, a hand posture refers to a 3D hand configuration, whereas a hand shape is the projected ....

....12 4. Application to Auslan Sign Recognition The functionality of the HMU tracker is evaluated by observing the tracking performance of static and dynamic signs. The kinematic data sequence attained by the 3D tracker is classified as a sign by the HMU classifier, an adaptive fuzzy expert system [11], 14] In the HMU system, a sign is represented by a combination of: a starting hand posture; motion information that describes the changes which occurred during the movement, such as the number of wiggles in a finger movement; an ending hand posture. The starting and ending hand ....

E. J. Holden,Visual Recognition of Hand Motion, PhD thesis, University of Western Australia, 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC