Figure recognizing system

• US 4,484,348 A
• Filed: 12/18/1981
• Issued: 11/20/1984
• Est. Priority Date: 12/22/1980
• Status: Expired due to Term

• Current Assignee: Tokyo Shibaura Denki Kabushiki Kaisha (Toshiba Corporation)
• Original Assignee: Tokyo Shibaura Denki Kabushiki Kaisha (Toshiba Corporation)
• Inventors: Shizuno, Masaaki
• Primary Examiner(s): Britton, Howard W.
• Application Number:
  US06/332,297
• Time in Patent Office: 1,068 Days
• Field of Search: 382/1, 382/30, 382/33, 382/34, 382/42, 382/48
• US Class Current: 382/102
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data

Abstract

In a figure recognizing system according to the invention comprising a scanning section, a detecting section, thesauruses, a similarity computing section, a storage register for storing detected information from the detecting section and a recognizing section for recognizing the final letter information, threshold table means in which letter combinations and corresponding threshold levels are stored as pairs, a category judgement section for receiving the output of the similarity computing section, obtaining a similarity difference from the maximum and next maximum similarity values and comparing the similarity difference thus obtained and a corresponding threshold level in the threshold table means to judge the category of said letter, format table means where a predetermined category series is stored, and format judging means for comparing the category supplied from the detecting section and category judgement section connected thereto and the category series supplied from the format table means.

Claims

1. A figure recognizing system comprising:

(a) a scanning section for reading out letter information of various letter species recorded on a recording medium;

(b) a detecting section connected to said scanning section for effecting letter detection from the letter information from said scanning section;

(c) thesauruses consisting of standard letter patterns of various letter species corresponding to said various letter species;

(d) a similarity computing section for receiving information from said detecting section and standard letter pattern information from said thesauruses and computing the similarity corresponding to the letter information from said detecting section;

(e) threshold table means in which letter combinations and threshold levels for the respective letter combinations are stored as pairs;

(f) a category judgement section for receiving the output of said similarity computing section, obtaining a similarity difference from the maximum and next maximum values of said similarity and comparing said similarity difference and a corresponding threshold level of said threshold table means to judge the category of said letter;

(g) format table means where a predetermined category series is stored;

(h) format judging means connected to said detecting section and category judgement section for comparing a category supplied from said detecting section and a category series supplied from said detecting section;

(i) storage means for storing the detected information from said detecting section; and

(j) a recognizing section connected to said format judging section and storage means for recognizing the content of said storage means to provide final letter information if a coincidence is obtained as a result of judgement in said format judging section.

2. The figure recognizing system according to claim 1, wherein said category judgement section includes:

a first comparing circuit for comparing the similarities provided from said similarity computing section to derive the maximum similarity (S_N1, S_A1) and the letter name (C_N1, C_A1) thereof and compute high and low values (S₁, S₂) of said maximum similarity (S_N1, S_A1) and letter names (C₁, C₂) corresponding to said high and low values;

a maximum value reserving register for storing the high and low values (S₁, S₂) output from said first comparing circuit and corresponding lettr names (C₁, C₂);

a subtracting circuit receiving the output of said maximum value reserving register and subtracting said low value (S₁) from said high value (S₂);

a retreaving circuit for retreaving the output of said maximum value reserving register and finding out the letter name combination coincident with said combination of letter names (C₁, C₂) to provide the corresponding threshold level (T_H);

a second comparing circuit for receiving the output of said subtracting circuit and the output of said retreaving circuit and comparing said subtracted value (S₁ -S₂) and said threshold level (T_H) to provide the category name to which C₁ belongs if S₁ -S₂ ≧

T_H and provide both the category name to which C₁ belongs and category name to which C₂ belongs if S₁ -S-hd 2<

T_H ; and

a category storage memory for storing a category name provided from said second comparing circuit.

3. The figure recognizing system according to claim 1, wherein said format judgement section includes:

a format selecting circuit for receiving the block position data supplied from said detecting section and selecting only the formats permisible in the block position of the received data as a category series; and

a format comparing circuit for comparing the n-th category of the category series supplied from said format selecting circuit and the n-th letter category supplied from said category judgement section and supplying the block and formats to said recognizing section if a coincidence is obtained while discarding the block if no coincidence is obtained.