Open AccessSpelled sign word recognition using key frame
Author(s) -
Rokade Rajeshree S.,
Doye Dharmpal D.
Publication year - 2015
Publication title -
iet image processing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.401
H-Index - 45
eISSN - 1751-9667
pISSN - 1751-9659
DOI - 10.1049/iet-ipr.2012.0691
Subject(s) - key frame , computer science , gesture , gesture recognition , sign language , artificial intelligence , key (lock) , frame (networking) , feature extraction , word (group theory) , speech recognition , computer vision , segmentation , sign (mathematics) , pattern recognition (psychology) , feature (linguistics) , process (computing) , invariant (physics) , mathematics , telecommunications , mathematical analysis , linguistics , philosophy , geometry , computer security , mathematical physics , operating system
In this study, the authors present a new system for sign language hand gesture recognition. Using video input, the system can recognise any spelled word or alphabetic sequence signed in American Sign Language. The three main steps in the recognition process include detection of the region of interest (the hands), detection of key frames and recognition of gestures from these key frames. The proposed segmentation algorithm distinguishes regions of interest from both uniform and non‐uniform backgrounds with an efficiency of 95%. The proposed key frame detection algorithm achieves an efficiency of 96.50%. A rotation‐invariant algorithm for feature extraction is additionally proposed, which provides an overall gesture recognition efficiency of 84.2%.