Character Reader

a device for the automatic identification of characters—that is, letters, numerals, or other symbols—that are printed or written on paper in the usual form by a human being. Character readers are intended mainly for automatic data input into computers or other automatic systems that process data directly from a sheet without the preliminary rerecording of the data on information carriers, for example, punched cards.

In character readers, character recognition is based on the following two procedures: the measurement of the “blackness”—that is, the absorption coefficient—of individual tiny (for example, 0.1 x 0.1 mm²) squares, into which a field containing a character to be read is divided, and a subsequent comparison of the results obtained against standard characters, which are idealized and generalized. As a rule, exact agreement between a written or printed character and a standard character is not necessary, since the comparison is usually continued until the least permissible value of a quantity characterizing the similarity of the written or printed character and the standard character is reached. As a result of the comparison, a code is generated that corresponds to the number of a standard character, the name of a character, or the position of a letter in an alphabet. The generated codes are usually obtained as electrical signals at the output of the reader.

Blackness is measured by using either a scanning system similar to that employed in a television camera or a photocell matrix, in which the blackness of the many tiny squares that make up a character is measured simultaneously by means of miniature photocells, for example, photodiodes. In its design, a photocell matrix resembles the retina of the eye.

In contrast to both a television camera and a facsimile apparatus, a character reader not only converts a visual image into an electrical signal but also rejects signals that correspond to unrecognizable characters, discards insignificant elements, and extracts the most important information concerning the assignment of a character being read to a specific class of characters. The simplest character readers are designed to read stylized characters, which are given a special form. An example of such characters is the numerals of the postal codes that are used on envelopes and postcards in the USSR. More complicated character readers are used to identify the font of a standard typewriter. However, the presence in an alphabet of letters with a similar outline—for example, the Russian letters ш and щ or э and 3—and the poor quality of characters printed with a typewriter make it difficult to obtain a high reliability in the automatic identification of typewriting.

The most advanced character readers are multifont readers, which are designed to read texts printed in various fonts or typewriter fonts. A multifont reader consists of the following components: an on-line storage device, in which standard characters of one or two fonts are stored; an auxiliary storage device for the storage of up to several dozen other fonts; an arithmetic unit, in which a character to be identified is compared with a standard character; and a program-control unit. In complexity, a multifont reader is comparable to a medium-capacity general-purpose computer. With time expended on such operations as the movement of a document and line searching taken into account, the reading rate of a multifont reader may be as high as several hundred characters per second.

As of 1976, experimental models of character readers for the identification of handwritten characters, mainly stylized numerals, had been developed. In such character readers, various methods of analyzing the geometric structure of a character are used rather than the method of direct comparison with a standard character.

Character readers are used in the processing of, for example, bank checks, bills, statements, bills of lading, supply orders, and statistical reports.