| 摘要 |
<p>A method and a system for recognizing a character with a hippocampal neural network learning algorithm are provided to maximize a character recognition ratio by recognizing the character from an image with the hippocampal neural network learning algorithm capable of increasing a recognition database regardless of the number of character recognition patterns/features. A scanning unit(20) forms a document image by scanning a document. A recognition unit(30) outputs a recognition result by receiving the document image from the scanning unit, extracting a character image of target characters from the document image, preprocessing/converting the extracted character image into text, and extracting features of the characters from the converted text. The recognition unit comprises a short-term recognition database(36) storing patterns having a value under a threshold set by an output frequency of the pattern, a long-term database(38) storing the patterns having the value over the threshold, and a processor(32). The processor outputs the recognition result by operating a recognition program(34), obtaining the pattern matched with the features of the extracted character with the short-term and long-term database.</p> |
