期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2015
卷号：6
期号：1
页码：289-293
出版社：TechScience Publications
摘要：There exists significant variation between sign language recognition processes across the world, although there are many similarities. Pre-processing, feature extraction and classification are the three major steps involved in the sign language recognition process. An analysis of scientific literature indicates the potential of various methods in achieving significantly high accuracy in image recognition. Further examination of the literature indicates the voluminous works carried out in American Sign Language recognition systems and most of these works compare the potential of various methods and combination of methods for their accuracy. Although, the comparison using randomly selected gestures for their potential would result in realistic overall accuracy for ASL where the gestures are simple and distinct, the complete adoption of such methods for Indian Sign Language (ISL) recognition may not be ideal due to the complexity in ISL. Other than static gestures, the dynamic gestures, gestures including facial expression, similarity in gestures, all increase the complexity of ISL. Therefore, the potential of different methods and their combinations need to evaluate in the context of ISL. A preliminary study to analyse the potential of promising feature extraction methods indicated that the methods could vary significantly while handling gestures with resemblances. This clearly indicates the necessity to screen gesture recognition methods for their accuracy in handling gestures in the context of complex ISL.
