A fault tolerant Chinese bank check recognition system based on SOM neural networks

The high-accuracy requirement excludes the application of common optical character recognition (OCR) technology. In this paper, according to the characteristics of the bank check, a fault tolerant Chinese bank check recognition system is presented and analyzed in detail. First, a high performance SOM classifier is adopted as the classifier to the numerals in the bank check. Next, a simple method is presented of SOM classifier to provide second recognition choice of the input numerals. We also propose a new idea of fault-tolerant recognition in the paper. With the combination of SOM classifier and fault tolerant technique, this system can reach high recognition rate and high reliability simultaneously in automatic processing of bank checks with the dynamic cipher code. A practical scheme of fault-tolerant recognition of bank check is described