This paper describes a continuous recognition method of sign language (numerals 0∼9 and a word [bad]) on hand motion by image processing technique. Finger Language communicates by means of state of the fingers and motion of the hand. Therefore, judgment rules are made clear from the state of fingers (bent, stretched, direction, angle of the interdigital web) for discriminating numerals. Each input hand image at 0.2(sec)is discriminated by using the judgment rules from the hand contour feature. Finger language can be recognized based on the continuous discrimination rule and above-mentioned judgment rules in real time.