Bangladeshi hand sign language recognition from video

Hand sign recognition has become a significant research topic as it requires voiceless communication with sign languages in many circumstances. In Bangladesh, some system has been established for hand signs recognition for bangle alphabets. In case of expressing the feelings of deaf and dumb people there are no alternative to words and sentences. In this regard framework has proposed a system for recognizing the Bangladeshi hand sign language of words and sentences and translating the signs into text with static hand images which have been generated from input videos. In our proposed system, RGB images which contain hand signs are firstly converted into YCbCr color space for skin color segmentation. After that hand portions which are our region of interest (ROI) have been obtained by eliminating other skin portions and Local Binary Pattern (LBP) has been applied on it for feature extraction. Finally, for classifying candidate features Support Vector Machine (SVM) is performed. To demonstrate the total effectiveness of this method experimental result have compared with other existing work related to Bangladeshi hand sign language. The experimental results denote that the proposed method outperforms the existing works. The system has achieved recognition accuracy of 94.26% for words and 94.49% for sentences.