CONTINUOUS RECOGNITION METHOD FOR JAPANESE FINGER LETTERS USING CONTOUR FEATURE

This paper deals with a continuous recognition method for finger signs used to express some numerals in Japanese sign language by image proc essing technology. There have been some reports about recognition meth ods for finger letters; however, these were concerned with the finger letters which could be recognized from only the hand shape. Some finge r letters are expressed by both hand shape and motion; therefore, it i s necessary to use sequential hand images to recognize such finger let ters. We propose a new continuous recognition method for sequential ha nd images and their contour curves. First, we define some properties o f the hand image with its contour curve. Next, we determine a judgemen t rule for discrimination of the hand shapes, based on the above prope rties. Finally, we describe a recognition rule to apply the judgement rule to some sequential hand images, and show some examples to demonst rate the feasibility of the continuous recognition method.