Citation Context

...nt end in several gesture recognition applications that take the hand configuration as input. Current systems rely almost exclusively on non-vision techniques to obtain such data, such as CyberGloves =-=[7, 18, 19, 30]-=- and color markers [11]. An automated computer vision technique like ours imposes no restrictions on users. It can also be used in domains where we do not have control of the data collection, and ther...

Citation Context

...anguage recognition, virtual reality, biometrics, and motion capture. Currently, systems requiring accurate 3D hand parameters tend to use magnetic tracking devices and other non vision-based methods =-=[15, 16, 20]-=-. Computer vision systems that estimate 3D hand pose do it only in the context of tracking [18, 7, 27, 21] . In that context, the pose can be estimated at the current frame as long as the system knows...

Citation Context

... number of approaches have been proposed for sign language recognition (see [14] for a recent review). Many approaches are not vision-based, but instead use magnetic trackers and sensor gloves, e.g., =-=[8, 11, 16, 22, 23, 25]-=-. Such methods achieve good recognition results on continuous Chinese Sign Language with vocabularies of about 5,000 signs [8, 23, 25]. On the other hand, vision-based methods, e.g., [2, 5, 6, 7, 9, 1...

Citation Context

...anguage recognition, virtual reality, biometrics, and motion capture. Currently, systems requiring accurate 3D hand parameters tend to use magnetic tracking devices and other non vision-based methods =-=[8, 9, 13]-=-. Computer vision systems that estimate 3D hand pose do it only in the context of tracking [11, 4, 17, 14] . In that context, the pose can be estimated at the current frame as long as the system knows...

Citation Context

... build HMMs. We also seek to find subunits but at sentence level. Also instead of using HMMs that are good for discretized state space representations, we use a continuous state space and a time-normalized representation. Most of the works in continuous sign language recognition have avoided the basic problem of segmentation and tracking of hands by using wearable devices, such as colored gloves, or magnetic markers, to directly get the location features. For example Vogler and Metaxas [9] have used 3D magnetic tracking system; Starner and Pentland [8] have used colored gloves while Ma et al. [5, 10] have used Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05) 1063-6919/05 $20.00 © 2005 IEEE Cybergloves. Bauer and Krass [1] use colored gloves for feature extraction. Fang et al. [3] used cybergloves and 3D position trackers. Only recently there has been effort to extract information and to track directly from color images, without the use of special devices [12], but it has only been used for isolated sign recognition. In this paper, we use only the plain color images, without using any augmenting wearable devices, for sentence lev...

Citation Context

...roblems by employing direct instrumentation of hands and body of signers, usually employing data gloves or similar equipment, coupled to magnetic position sensors (e.g. [VM98, VM01], [Kad98b, Kad95], =-=[MGWW00]-=-, [FGC + 01], [WGX01]). These devices give direct (but noisy) readings of the position and orientation of body parts, as well as the flexion of fingers. Despite measurement problems that might be rela...

Citation Context

...approaches have been proposed for sign language recognition (see [18] for a recent review). Many approaches are not vision-based, but instead use input from magnetic trackers and sensor gloves, e.g., =-=[12, 16, 23, 29, 30, 32]-=-. Such methods achieve good recognition results on continuous Chinese Sign Language with vocabularies of about 5,000 signs [12, 30, 32]. On the other hand, vision-based methods, e.g., [3, 7, 10, 11, 1...

Citation Context

...that, unlike many of the previous works in sign language that perform tracking of the hands using 3D magnetic trackers or color gloves (Fang et al., 2004; Vogler and Metaxas, 2001; Wang et al., 2002; =-=Ma et al., 2000-=-; Cooper and Bowden, 2009), our representation does not require tracking and relies on skin segmentation. We present a Bayesian framework to extract the common subsequences or signemes from all the gi...

Citation Context

...ber of approaches have been proposed in the literature (see [13] for a recent review). Many approaches are not vision-based, but instead use input from magnetic trackers and sensor gloves, e.g., [9], =-=[12]-=-, [16], [20], [21], [23]. Such methods achieve good recognition results on continuous Chinese Sign Language with vocabularies of about 5,000 signs [9], [21], [23]. On the other hand, vision-based meth...