Collaborative learning system and pattern recognition method

• US 6,385,339 B1
• Filed: 06/30/2000
• Issued: 05/07/2002
• Est. Priority Date: 09/14/1994
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Miura, Masaki, Yokota, Toshimi, Arai, Toshifumi, Shojima, Hiroshi, Fukunaga, Yasushi, Kuzunuki, Soshiro, Gunji, Keiko
• Primary Examiner(s): Patel, Jayanti K.
• Application Number:
  US09/609,326
• Time in Patent Office: 676 Days
• Field of Search: 382/100, 382/141, 382/159, 382/162, 382/165, 382/181, 382/182, 382/190, 382/198, 382/199, 382/185-189, 382/203, 382/219, 382/309, 345/762, 345/767, 345/751, 706/20, 707/2, 707/530, 707/532, 704/10
• US Class Current: 382/181
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 30/242 Division of the character sequences into groups prior to recognition; Selection of dictionaries

Abstract

The present invention relates to a multi-user pattern data processing system configured for various kinds of patterns, such as characters, to learn efficiently and effectively. The present invention comprises an input arrangement 1 having patterns entered therein from a plurality of users, a dictionary 2 having patterns and attributes of the patterns belonging thereto defined therein, a recognizing arrangement 4 that receives the entered pattern and a group attribute and retrieves from among the patterns entered in the dictionary and having a corresponding group attribute to feed out the category, and a dictionary editing arrangement 5 for extracting a pattern used in common by a group before editing the pattern, such as entering it in the common dictionary, with the group having the user made to belong thereto. The invention allows the dictionary used to the pattern recognition to learn very efficiently and effectively.

Claims

1. A pattern data processing system, comprising:

a plurality of input arrangements for receiving given patterns at least;

a plurality of dictionaries in which data classified plurally for a plurality of users are registered and each of which has a group attribute of the users therein;

a common dictionary;

a dictionary reference arrangement adapted to read out a code corresponding to the patterns received from said input arrangements with reference to the plurality of dictionaries in an order on the basis of priority level set in advance for the plurality of dictionaries; and

a dictionary editing arrangement which edits respective contents in the dictionaries on the basis of group attributes of the users set in the respective dictionaries,

wherein, if the pattern given through any of the input arrangements meets predetermined conditions the dictionary editing arrangement edits the data in the dictionaries and enters the pattern and code corresponding to the pattern in the common dictionary, and the users having the same group attribute are accessible to the common dictionary.

2. The pattern data processing system according to claim 1 wherein the group attribute of the users is a name of a group having a plurality of the users made to belong thereto as members.

3. The pattern data processing system according to claim 1 wherein the predetermined conditions are that the patterns given through the input arrangements by two or more users having the same group attribute are the ones having a predetermined relationship corresponding to the same code and that the given patterns and the codes corresponding to the patterns are classified and entered in the dictionary as data belonging to the users.

4. The pattern data processing system according to claim 3 wherein the predetermined relationship is that a calculated value of a matching function for between the patterns is not lower than a predetermined threshold value.

5. The pattern data processing system according to claim 1 wherein the patterns include at least one of hand-written kanji, hiragana, alphabets, numerals and symbols.

6. The pattern data processing system according to claim 1 wherein the data edition made by the dictionary editing arrangement is to add or delete patterns in the dictionary.

7. The pattern data processing system according to claim 1 wherein the dictionary further has data of the group attributes of the patterns added to the patterns entered in the dictionary and the dictionary editing arrangement updates the group attributes.

8. A pattern data processing system, comprising:

a plurality of input arrangements for receiving given patterns at least,

a dictionary having data classified plurally for a plurality of users,

a common dictionary, and a dictionary editing arrangement,

wherein the dictionary has a group attribute of the users therein, if the pattern given through any of the input arrangements meets predetermined conditions the dictionary editing arrangement edits the data in the dictionary and enters the pattern and code corresponding to the pattern in the common dictionary, and the users having the same group attribute are accessible to the common dictionary and

wherein the dictionary editing arrangement further extracts parts of two or more patterns being entered in the dictionary and having the same code to make cross-over, or operate as a kind of genetic algorithm, of a pattern to generate and adds the generated pattern to at least either of the dictionary or the common dictionary.

9. The pattern data processing system according to claim 8 wherein the dictionary editing arrangement further assigns adaptabilities to the patterns entered in the dictionary, the pattern being processed at high frequency as the adaptability having high value, and the patterns are selected in descending order of adaptability to cross over.

10. A pattern recognizing system, having an input arrangement for entering a pattern, an approximating arrangement for approximating the pattern with some elements, a dictionary for having dictionary pattern approximated with some element stored therein in advance, and a distance calculating arrangement for comparing the approximated entered pattern with the approximated dictionary pattern, wherein the distance calculating arrangement comprising:

a range correspondence arrangement for making each element of the approximated entered pattern to correspond the respective

elements of the approximated dictionary pattern, and

a particular shape difference calculating arrangement for calculating a particular shape difference for evaluation of whether or not a particular shape formed of some elements of approximation results exists at a corresponding position of the dictionary pattern an entered pattern.

11. A pattern recognizing system according to claim 10 wherein the approximating arrangement approximates continuous parts of the pattern with some broken lines.

12. A pattern recognizing system according to claim 10 wherein the approximating arrangement is capable of leaving or knowing by analogy a particular shape of the pattern before the approximation.

13. An information processing system having a plurality of information processors connected via a signal line, each of said plurality of information processors comprising:

an input section for entering characters;

a dictionary for storing a plurality of characters, a plurality of character codes associated therewith and a plurality of different kinds of character codes thereof;

a memory section for storing a priority level of dictionaries in said plurality of information processors connected to said signal line; and

a dictionary reference section adapted to read out a different kind of character code associated with said character from its own dictionary or from at least one other dictionary in another information processor on the basis of said priority level stored in said memory section, for outputting said different kind of character codes, upon coincidence with said character having been read.