Discriminator between handwritten and machine-printed characters

A discriminator comprising:

• scanning means for optically scanning a predetermined region including characters recorded on a document in a predetermined scanning width and converting video image thereof into electric signals;
• sampling means for sampling said electric signals with a clock signal of a predetermined frequency and outputting the sampled signal as pixel signals;
• memory means for temporarily storing said pixel signals;
• means for successively reading out from said memory means unit region signals each consisting of m lines by n rows of pixel signals as unit pattern signals, m and n representing natural numbers;
• means for extracting horizontal stroke, vertical stroke or slanted stroke component of the characters on the basis of pixel signals contained in said unit pattern signal;
• means for calculating the ratio of the number of the slanted stroke components to the total number of the horizontal, vertical and slanted stroke components extracted from the unit pattern signals for said predetermined region;
• and
• means for discriminating said characters as handwritten ones if said ratio is above a predetermined threshold or as machine-printed ones if it is not above the threshold.

A handwritten character discriminating apparatus comprising:

• a register for storing a video image of a predetermined region containing characters read out by a photoelectric converter-scanner;
• means for picking up from said register subregion pixel signals of said video image as unit signals;
• a pattern memory for storing predetermined pattern signals of horizontal, vertical and slanted stroke lines;
• comparator means for comparing said unit signals with pattern signals read out from said pattern memory and generating an identity signal for each of unit signals found identical with one of said pattern signals;
• counter means for counting the number of generation of the identity signals for each stroke line generated from said comparator means;
• arithmetic means for receiving the outputs of said counter means and calculating the ratio of the number of generation of the identity signals for said slanted stroke line to the total number of generation of the identity signals for said horizontal, vertical and slanted stroke lines;
• and
• discriminating means for discriminating said characters as handwritten ones if said ratio is greater than a predetermined threshold.

A method for identifying handwritten characters and machine-printed characters comprising the steps of:

• converting a video image of a predetermined region containing characters on a document into electric signals;
• picking up from said electric signals unit signals corresponding to subregions obtained by dividing said predetermined region;
• detecting horizontal, vertical or slanted stroke components included in each of said subregions from said unit signals;
• counting numbers of the detections of the horizontal, vertical and slanted stroke components, respectively;
• calculating a ratio of the counted numbers of the slanted stroke component to the total of the horizontal, vertical and slanted stroke components;
• and
• judging said characters on the document as handwritten ones and machine-printed ones if said ratio is above and not above a predetermined threshold, respectively.