Abstract: This paper presents a system for the recognition of sign language based on a theory of shape representation using size functions proposed by P. Frosini [5]. Our system consists of three modules: feature extraction, sign representation and sign recognition. The first performs an edge detection operation, the second uses size functions and inertia moments to represent hand signs, and the last uses a neural network to recognize hand gestures. Sign representation is an important step which we will deal with. Unlike previous work [15, 16], a new approach to the representation of hand gestures is proposed, based on size functions. Each sign is represented by means of a feature vector computed from a new pair of moment-based size functions. The work reported here indicates that moment-based size functions can be effectively used for the recognition of sign language even in the presence of shape changes due to differences in hands, position, style of signing, and viewpoint.