Recognition of Isolated Signs in Ethiopian Sign Language
No Thumbnail Available
Date
2014-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Sign Language is a visual gestural language which is used by deaf people for the purpose of
communication. Even if it is widely used in the deaf community, most of the hearing people do
not understand it. Due to this communication gap deaf people encounter so many problems in
their daily life since they are living with the people who communicate with spoken languages. To
narrow this communication gap there should be technological solutions that assist the deaf
community. Sign language researches are striving to fill the gap of the communication. The goal
of this research is the recognition of Isolated Signs in Ethiopian Sign Language.
The proposed system accepts videos of Isolated signs and get frames in the videos. On each
frame, skin color detection algorithm is applied and the equivalent binary image is produced
which has white value for skin color and black value for other region. Based on the detected skin
regions the hands and head are segmented from other parts of the body since they have very
important role in the signing process. Important features are extracted from the segmented body
parts and these values are converted into symbols using k-means algorithm. Hidden Markov
Models trained using these symbols and Baum Welch algorithm and stored in the database. A
Viterbi decoding is applied in the recognition process using the trained HMMs and symbol
sequences of the testing Signs which is prepared by the above process.
We evaluated our Isolated Sign Language recognition system and we found a recognition rate of
86.9% using 8 features and 83.5% using only 3 features.
Keywords
Sign Language, Skin Detection, Hidden Markov Model, Recognition
Description
Keywords
Sign Language, Skin Detection, Hidden Markov Model, Recognition