LIBRAS translator via Web for Mobile Devices

This paper presents the development of an application dedicated to the translation of gestures of LIBRAS (Brazilian language of signs) via Web for Mobile Devices. Such application was directed for support to mobile devices, such as, Tablets, cellphones and Palms, and its use was applied in a Linux platform. Its operation consists in the capture, analysis and identification of the signs of alphabet of the LIBRAS, realized by users. PDI (Digital Image Processing), Web Service and Computational Intelligence were employed in the development of the project. The process starts from the identification of the captured images, where occurs the background subtraction, which is implemented to split the acquired images in two subsets: the background and interest objects(foreground). After this division, the representation of the interest objects is realized using an image description technique, known as Calculation of Moments Invariants. Then these objects are sent to the web server, analyzed and classified, through Computational Intelligence. At the end of process, the letter corresponding from the alphabet is obtained in the mobile device. For development and tests, the mobile device Nokia Internet Tablet N810 and the language Python are utilized. Part of the solution executes in one server and was developed using the language Java with XML/RPC (Remote Procedure Call) as alternative in relation to the limitation of the processing in the mobile device.