Feature Extraction in Arabic Sign Language using Hand and Wrist Localization Techniques

Recognition of hand and wrist gestures is the vital and challenging problems in human computer interaction (HCI) community. In this paper, an essential phase in the process of identifying hand gestures is presented using innovative hand detection and wrist localization method. There has not been much focused to localizing the wrist, and available work are very few which cover a wide range of circumstances. On a publicly available dataset, the suggested approach was implemented, tested and the findings obtained support its effectiveness. By contrasting this method with earlier research, the effectiveness of this method and the significance of the localization with wrist phase has been presented. In this work, we suggest to use proposed method which is applicable for general sign language recognition (SLR) and Arabic digit SLR. The proposed technique (having error = 125 when maximum detection (E) = 1 and error 175 when E = 0.5) exhibits the better results as compeered to the existing method (having error = 131.8 when E = 1 and error 323.3 when E = 0.5). Using this paper, the researchers will explore to investigate novel creation of a real-time Arabic SL application using hand characteristics in order to propose other classifiers to boost the effectiveness of the detection of hands and wrists method.