Research On Construction Of Transitional Video Library Oriented To Sign Language Video Splicing

In China, there are lots of deaf people, sign language is their first language, andmain way to communicate with others in daily life. Along with progress of humansociety and development of computer science, synthesized Chinese sign languageexpression may provide more convenience for deaf to integrate into the society. Thesign language synthesis based on video splicing has a strong sense of realism andacceptability. At first, the collected sign language video data is divided into severalshort segments according to certain grammatical rules or semantic. In the synthesisstage, the video clips are reorganization. Words from different sentences have thehand trajectory, gestures and other differences, so the result video is not smoothenough without transition frames sequence. So building transitional videos library isan important basic work for sign language synthesis based on video splicing.In this paper, the goal is to research transitional video library constructionmethod oriented to sign language video splicing, and the main contents are as follows:1. Human joints tracking based on improved particle filter tracking algorithm. Annew improved particle filter algorithm is proposed in this paper, whichintroducing optical flow as speed characteristics, and wavelet moment asprofile feature, to improve the accuracy of tracking body joints position. Thusobtaining more accurate of the left elbow, right elbow, left wrist and right wristmotion information is as a result.2. Based on the body joints motion information，using information fusion method,the transition point frame boundary judgments, in order to build transitionalvideo library.