|
Acadêmico(a): Júlio César Batista |
Título: SIGNA: UMA APLICAÇÃO PARA ENSINO-APRENDIZAGEM DA LIBRAS |
|
Abstract: |
This work presents the development of an application to assist in teaching and learning the Brazilian Sign Language using the Leap Motion device. The sign recognition is done in real time using machine learning algorithms to classify features extracted from the user of hands with the help of Leap Motion. These algorithms use an initial sample database that can be expanded by the users. The Accord.NET framework was used in order to provide implementations of the machine learning algorithms and the real time recognition was done using WebSocket with the ASP.NET SignalR framework. Tests were performed to verify the performance and usability of the application. From performance tests it was possible to notice that the application can be used for sign recognition in real time, and crossed validation tests showed 86% of accuracy in sign recognition. From usability tests it was possible to observe that the application can be used to help the basic teaching and learning of the Brazilian Sign Language. However, the application has limitations with signs that exhibit similar features, facial expression and when they show occlusion of fingers for the Leap Motion. |
|
|