Sample DATASET

Reliable hand gesture recognition is an important problem for automatic sign language recognition for the people with hearing and speech disabilities. For this purpose, we have created a new benchmark database of multi-oriented, isolated, static ASL digit and alphabet images using depth sensor Microsoft Kinect V2. Further, we design a statistical-geometrical feature combination for recognition of the hand gestures out of our dataset. For comparison purpose, we have used public datasets collected from the same sensor and found that accuracy of our system is comparable and sometimes higher than the state-of-the-art methods reported so far. We have also collected data for JU_V2_DIGIT dataset from 0 to 9 and for JU V2 ALPHA dataset from a to z. We achieve an accuracy of 96.10%, 95.75% and 95.65% for Digits, alphabets and alphanumerals accordingly, which is first of its kind.