The major challenges that sign language recognition (SLR) now faces are developing methods solve large-vocabulary continuous problems. In this paper, transition-movement models (TMMs) proposed to handle transition parts between two adjacent signs in SLR. For tackling mass movements arisen from a large vocabulary size, temporal clustering algorithm improved k-means by using dynamic time warping as its distance measure is dynamically cluster them; then, an iterative segmentation for...

The major difficulty for large vocabulary sign recognition lies in the huge search space due to a variety of recognized classes. How reduce time without loss accuracy is challenging issue. In this paper, fuzzy decision tree with heterogeneous classifiers proposed language recognition. As each feature has different discrimination gestures, corresponding are presented hierarchical attributes. A one- or two- handed classifier and hand-shaped little computational cost first used progressively...

The major challenges that sign language recognition (SLR) now faces are developing methods solve large vocabulary continuous problems. In this paper, SLR based on transition movement models is proposed. proposed method employs the temporal clustering algorithm to cluster a amount of movements, and then corresponding training also presented for automatically segmenting these models. clustered can improve generalization models, very suitable SLR. At last, estimated together with viewed as...

Unsupervised representation learning for dynamic graphs has attracted a lot of research attention in recent years. Compared with static graph, the graph is comprehensive embodiment both intrinsic stable characteristics nodes and time-related preference. However, existing methods generally mix these two types information into single space, which may lead to poor explanation, less robustness, limited ability when applied different downstream tasks. To solve above problems, this paper, we...

In this paper, a vision-based medium vocabulary Chinese sign language recognition (SLR) system is presented. The proposed consists of two modules. the first module, techniques robust hands detection, background subtraction and pupils detection are efficiently combined to precisely extract feature information with aid simple colored gloves in unconstrained environment. Meanwhile, an effective efficient hierarchical description scheme different scale features characterize proposed, where...

The major difficulty for large vocabulary sign language or gesture recognition lies in the huge search space due to a variety of recognized classes. How reduce time without loss accuracy is challenge issue. In this paper, hierarchical decision tree first presented based on divide-and-conquer principle. As each feature has different importance gestures, corresponding classifiers are proposed attributes. One- two- handed classifier with little computational cost used eliminate many impossible...

The aim of sign language recognition is to provide an efficient and accurate mechanism transcribe into text or speech. State-of-the-art should be able solve the signer-independent problem for practical application. In this paper, a hybrid SOFM/HMM system, which combines self-organizing feature maps (SOFMs) with hidden Markov models (HMMs), presented Chinese recognition. We implement system. Meanwhile, results from HMM-based system are provided as comparison. Experimental show increases...

In sign language recognition, using subwords instead of whole signs as basic units scales well with increasing vocabulary size. However, there are no defined in the signs' lexical forms. How to automatically extract is a challenging issue. this paper, novel approach proposed these from Chinese (CSL). Signs can be broken down into several segments hidden Markov models which each state represents one segment. Temporal clustering algorithm presented segments. The 238 extracted 5113 signs, and...

Sign language recognition is to provide an efficient and accurate mechanism transcribe sign into text or speech. State-of-the-art should be able solve the signer-independent continuous problem for practical applications. A divide-and-conquer approach, which takes of Chinese Language (CSL) as subproblems isolated CSL recognition, presented recognition. In proposed improved simple recurrent network (SRN) used segment CSL. The outputs SRN are regarded states hidden Markov models (HMM) in...

In sign language recognition, using subwords instead of whole signs as basic units scales well with increasing vocabulary size. However, there are no defined in the signs' lexical forms. How to automatically extract is a challenging issue. this paper, novel approach proposed these from Chinese (CSL). Signs can be broken down into several segments hidden Markov models which each state represents one segment. Temporal clustering algorithm presented segments. The 238 extracted 5113 signs, and...

Using abundant Web resources to mine Chinese term translations can be applied in many fields such as reading/writing assistant, machine translation and cross-language information retrieval. In mining English of terms, how obtain effective pages evaluate candidates are two challenging issues. this paper, the approach based on semantic prediction is first proposed pages. The method predicts possible meanings according each constituent unit term, expands these items using semantically relevant...

This paper presents a Chinese sign language/spoken language dialog system based on the technique of large vocabulary continuous recognition (SLR) and synthesis (SLS), which is new development for HandTalker. In SLR module, fuzzy decision tree with heterogeneous classifiers presented signer-independent SLR, then transition movement models proposed. SLS three key techniques: realistic 3D facial animation gesture retargeting synchronization modal lip motion are employed to improve vividness.

The major difficulty for large vocabulary sign language or gesture recognition lies in the huge search space due to a variety of recognized classes. How reduce time without loss accuracy is challenge issue. In this paper, hierarchical decision tree first presented based on divide-and-conquer principle. As each feature has different importance gestures, corresponding classifiers are proposed attributes. One- two- handed classifier with little computational cost used eliminate many impossible...

Social media, e.g. Weblog and Internet forum, generate rich historical textual datasets which record lots of valuable events. Automatic event detection tries to discover important interesting events their related documents. Existing solutions detection, however, are mostly proposed for high quality news stories may not work well when they applied noisy social media datasets, where content varies drastically from informative trivial or even spamming. In this paper, an framework, directly...

Unsupervised representation learning for dynamic graphs has attracted a lot of research attention in recent years. Compared with static graph, the graph is comprehensive embodiment both intrinsic stable characteristics nodes and time-related preference. However, existing methods generally mix these two types information into single space, which may lead to poor explanation, less robustness, limited ability when applied different downstream tasks. To solve above problems, this paper, we...