When Neural Networks Play Sherlock Holmes
The fast, reliable, and computerized classification and matching of fingerprint images is a remarkable problem in pattern recognition which has not yet received a complete solution. Automated fingerprint recognition systems could have an extremely wide range of applications, well beyond the traditional domains of criminal justice. Such systems could in principle be used in any situation where identification, verification, and/or access control are paramount. A few examples include all identification cards systems, such as driver licenses, computer security systems, entitlement systems (such as welfare), access control systems (for instance, in airports or hospitals), and credit card (as well as several other types of financial transactions) validation systems. Automated fingerprint recognition systems could also render the use of locks and keys obsolete and be installed in cars, homes, and hotels. Although it is beyond our scope to discuss
either the existence of real markets for such applications or their complex juridical implications, it seems intuitively clear that fingerprint recognition is a welldefined problem in pattern recognition within the reach of our current technology and which should be amenable to neural network techniques. Our purpose here is to give a brief account of our current results on the application of neural network ideas to the problem of fingerprint matching. In particular, we describe the architecture, training, and testing of a neural network algorithm which, when presented with two fingerprint images, outputs a probability p that the two images originate from the same finger.