This paper deals with the possibility of using ARTMAP neural network for searching fingerprint patterns from a large database. ARTMAP has the ability to perform concurrent processing, to learn fast, and to make decisions. Since ARTMAP learning is self-stabilizing, it can continue to learn from one or more databases, without performance degradation, until its full memory capacity is utilized. Generally, fingerprint matching is based on local ridge characteristics, and its efficiency depends on minutiae extraction. The proposed method uses only gray level values of the image pixels along with its neighboring ones, instead of ridge features. [ABSTRACT FROM AUTHOR]