【Mitsubishi Electric】 M2481 and M2483

In June 1976, Mitsubishi Electric announced the M2481 optical character reader that recognized handwritten characters entered in boxes on page forms. Later, in April 1979, the company announced the M2483 model, which had improved reading performance and was easier to use, since it eliminated many of the writing implement restrictions. The following description refers to the M2483. The specifications of both models are given after the description.

The M2483 optical character reader had the following features:

• Mitsubishi Electric developed a red fluorescent light with a peak spectrum emission at 660 nm (previously, incandescent lights were used). This light made it possible for the M2483 to read forms filled in with regular ballpoint pens with black or blue ink, black felt-tip pens, or fountain pens with black or blue-black ink, in addition to pencils, special OCR ballpoint pens, and other writing implements that could be used with the previous model. This flexibility expanded the model’s application scope to include accounting and management operations and application and registration operations that used photocopy paper.
• The paper-feed mechanism was improved to accept landscape-oriented A4 forms and forms with binding holes or sprocket holes. High-grade paper could also be used with running costs half of those for OCR paper. As with the previous model, the M2483 did not require forms to have alignment or timing marks.
• The M2483 could read up to four lines per second (compared to the one line per second on the M2481). The extra speed was gained from the optical scanning assembly’s stationary head with four 512-bit line image sensors and a two-line buffer memory.
• A pipeline process was used to recognize characters using four microprocessors (preprocessor 1 (P1), preprocessor 2 (P2), main processor (MP), and auxiliary processor (AP)), which improved the recognition performance.
  P1 cropped out, normalized, and smoothed characters. P2 converted the character images to binary data. MP was in charge of identifying characteristics and most of the decision process by referring to a recognition dictionary. It included multiplier logic to find sums and a dedicated curve-tracing circuit. AP made the final character decisions based on candidate characters forwarded from MP using the character size, its vertical position, and information about adjacent characters. P1 and AP were single-chip 8-bit CPUs, and P2 and MP were high-speed bit-slice microprocessors.
• Up to 100 format programs, which defined form read controls (read positions and character types) and read data processes (data checks and output devices), could be stored on one system disk, allowing 100 types of forms to be processed simultaneously.
• The M2483 provided rich error checking functions and processing modes to improve operability and the reliability of results.
  Error checking included field checks (character type, number of digits, etc.), calculation checks between fields, calculation checks between forms, and verification checks.
  In terms of processing modes, the M2483 had an immediate correction mode, which suspended reading operations when an error was discovered or an unreadable character was found, a batch correction mode, which continued reading while flagging errors and recording them in a buffer that was written to the output device when the reading operations finished, and a mode that ejected forms with errors or unreadable characters to a reject stacker and did not output the read results. It also had a mode that displayed the read results and allowed the operator to confirm and correct data even if there were no errors or unreadable characters.
  
  

The specifications of the M2483 optical character reader were as follows: