Abstract: A character recognition system is disclosed in which each row of a plurality of parallel rows of binary signals representing the configuration of an unknown character is sequentially compared to a plurality of templates, each representing a known character. Each of the rows of signals is compared to feature characteristics of the unknown character twice. The rows of signals are first shifted one signal position and compared. The signals are then compared in their original position to detect an unknown character which has been printed off-center. Only those rows of signals which are required to contain a feature characteristic of the unknown characters are compared. The resulting data is utilized by a data processor apparatus identifying the unknown character.

Abstract: A method and apparatus for selecting the start of an unknown character sensed by a character recognition system. A multi-channel read head develops a plurality of analog waveforms which form the image of an unknown character sensed by the read head. After the analog waveforms have been digitized and rectified, static and dynamic threshold values are applied producing a multi-column binary bit map of the unknown character. The start of the character is detected when two adjacent rows of the bit map are found to have five consecutive binary bits representing a character. The end of the character is detected when the rows of the binary bits representing background data have a width which is inversely proportional to the width of the character detected. This information is used by a character recognition apparatus in locating the start of the character and the data generated during a character recognition operation.

Abstract: A system for processing data used by a recognition system in recognizing an unknown character. A multi-channel read member generates a plurality of parallel rows of data bits which taken together form an image of the character to be recognized. In order to align the rows of data in their proper relationship, each row of data bits is examined to determine if the row contains data representing a feature characteristic of the unknown character or a clear channel. The top row of the character is selected as the first row of data bits representing a feature characteristic of the character which follows the largest number of consecutive clear channel rows. The top two rows of the character are then compared. The row having the greatest number of feature characteristic data bits is selected as the top row of the character. Signals identifying the top row of the character are generated for us by a recognition apparatus in recognizing the unknown character from the rows of data bits generated.