JPS57146380A - Address reader - Google Patents

Abstract

PURPOSE:To increase the performance of reading with a word recognition due to the redundancy of a word, by passing a limit of the reading performance per one character for a character reader through repetitive word recognition for multi- stage of an address consisting of consecutive word trains. CONSTITUTION:Numerals and words are discriminated for the size at a numeric code pickup section 3 and a word pickup section 2 with a numeral reading output 101 from a character reading section 1, alphabetical capital and small letter reading outputs 102 and 103, and a pre-processing information 104 such as character location and size, and a word train is stored at a word buffer 4. The word train stored in the buffer 4 is set to a large, a medium and a small section word buffers 6-8 by discriminating the address based on a postal number at an address constituting word pickup section 5. The words set in a large, a medium and a small district geographical name dictionary 10-12 and those from the buffers 6-8 are compared at an optimum coincidence detecting section 14 with the combined order, and the combination having the highest similarity from the result of comparison is discriminated at an address descriminating section 18 and they are given to a number discriminating section 22 via buffers 19, 20 and 21.