14 SBC Journal on 3D Interactive Systems, volume 4, number 1, 2013 An evaluation of real-time requirements for automatic sign language recognition using ANNs and HMMs - The LIBRAS use case Mauro dos Santos Anjo, Ednaldo Brigante Pizzolato, Sebastian Feuerstack Computer Science and Engineering Department Universidade Federal de Sao˜ Carlos - UFSCar Sao˜ Carlos - SP (Brazil) Emails:
[email protected],
[email protected],
[email protected] Abstract—Sign languages are the natural way Deafs use to hands in front of the body with or without movements communicate with other people. They have their own formal and with or without touching other body’s parts. Besides semantic definitions and syntactic rules and are composed by all the hand and face movements, it is also important to a large set of gestures involving hands and head. Automatic recognition of sign languages (ARSL) tries to recognize the notice that face expressions contribute to the meaning signs and translate them into a written language. ARSL is of the communication. They may represent, for instance, a challenging task as it involves background segmentation, an exclamation or interrogation mark. As sign languages hands and head posture modeling, recognition and tracking, are very different from spoken languages and it is hard temporal analysis and syntactic and semantic interpretation. to learn them, it is natural, therefore, to expect computer Moreover, when real-time requirements are considered, this task becomes even more challenging. In this paper, we systems to capture gestures to understand what someone present a study of real time requirements of automatic sign is feeling or trying to communicate with a specific sign language recognition of small sets of static and dynamic language in order to promote social inclusion.