Detail of publication

Abstract

In the paper we give a brief introduction into sign language recognition and present a particular research task, where the access to MetaCentrum computing facilities was highly beneficial. Although the problem of signed speech recognition is currently being researched into by many research institutions all around the world, it lacks of a generally accepted baseline parametrization method. Our team introduced a parametrization method based on skin-color detection and object tracking. Because of the relatively high amount of information that is produced during the parametrization process, a method that reduces the unnecessary information while keeping the necessary information is required. Such methods are called feature space dimension reduction methods. We used the MetaCentrum facilities to evaluate several methods used widely in the field of acoustic speech recognition and their influence on recognition score of a Czech Sign-Language recognizer.