Abstract
Artificial Neural Network is an information treatment theory based on the nerve circuit in biology. Since it has pattern learning function, many studies have been made and applied for various pattern recognition systems. lt has been reported that the theory can also be applied for paper money recognition. In this paper, a new approach is proposed to optimize the location of sensing lines that is required in the development of paper money recognition system with Artificial Neural Network. Up to present, there is no definition to quantify the ability of recognition, therefore, the location of sensing lines has been decided by try and error. Our approach consists of two processes: firstly, S/N ratio is used to express the robustness of recognition, and secondly, the selection process of the number and location of sensing lines was optimized. The correlation between the S/N ratios of different sensing lines was analysed using orthogonal arrays. Already circulated notes were used in the experiment as a noise factor. The results showed a good correspondence with theoretical values.
