A NEURAL-NETWORK-BASED MODEL FOR PAPER CURRENCY RECOGNITION AND VERIFICATION

This paper describes the neural-based recognition and verification tec hniques used in a banknote machine, recently implemented for accepting paper currency of different countries. The perception mechanism is ba sed on low-cost optoelectronic devices which produce a signal associat ed with the light refracted by the banknotes. The classification and v erification steps are carried out by a society of multilayer perceptro ns whose operation is properly scheduled by an external controlling al gorithm, which guarantees real-time implementation on a standard micro controller-based platform, The verification relies mainly on the prope rty of autoassociators to generate closed separation surfaces in the p attern space. The experimental results are very interesting, particula rly when considering that the recognition and verification steps are b ased on low-cost sensors.