Accés ràpid intranet

Més informació...

a a a

Deim Seminar


Sign Language Recognition and the Use of Bag of Visual Words for Recognizing Hand Configurations


Jaume Vergs Llah

Professor/a organitzador/a

Agust Solanas


Empresa CRIC


25-11-2010 16:00


Within the European Project SignSpeak, which aims at the automatic translation of sign language by means of computer vision and machine learning techniques, one central problem is that of detecting and recognizing both hand position and configuration. Unlike the similar the case of face detection and recognition, hands are a very deformable and variable object, which difficult its description in a automatic way. Besides, the movement of hands within a sign belonging to a specific sign language is usually swift and encompasses very few frames, where the shape of the signing hand tends to appear blurred, obscured by occlusions, cluttered and ill-defined. Part of our work in such project tried to overcome such problems and different general approaches to recognize objects have been applied to hands, being those based on Bag of Visual Words those which have obtained better and more promising results, at least, in the case of single hands in a limited vocabulary of hand configurations.


Laboratori 231