Sign language is one of the ways to communicate by not making sound, as it is communicated by using hands and expressions. Sign language does not get much attention from the public because it is used by the minority of the society, hence the limited resources available for learning them. This paper proposes a sign translation system called ReadMe that uses deep learning approach, specifically the Convolutional Neural Networks (CNNs) to train the recognition model. However, training ReadMe using the CNNs revealed a low accuracy of 39.8% due to small size of training dataset, hence testing is aggravating. In order to increase the recognition accuracy in the future, CNNs algorithm in ReadMe will be trained using a larger dataset from both American Sign Language (ASL) and Malaysian Sign Language (BIM). In addition, the system is also hoped to enable users to train new gestures. Only by means of crowdsourcing that the system able to expand its vocabulary without facing the knowledge engineering bottleneck.
Sign Language Translation System Using Convolutional Neural Networks Approach
Lect.Notes Mechanical Engineering
Advances in Mechatronics, Manufacturing, and Mechanical Engineering ; Kapitel : 17 ; 161-171
2020-08-06
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Real-Time Traffic Sign Recognition Using Convolutional Neural Networks
Springer Verlag | 2021
|Convolutional Neural Networks for Traffic Sign Recognition
TIBKAT | 2021
|