JPS60107191A - Character recognizing system - Google Patents

Abstract

PURPOSE:To attain reduction in mis-recognizing rate and high speed processing by inputting the same character to plural systems from a character string being a processing object at the same time so as to allow each system to apply shared recognizing processing in parallel. CONSTITUTION:An observation character generation processor 1 observes a character ''ASE'' (sweat in Japanese) and gives a character number and observed character information to buffers 13-15 respectively. A character recognizing processor 2 forms segments for the observed character information (''ASE'') of the buffer 13, and divides the character into character information of a left- hand radical and a right-hand radical. The are added with the character number and the left-hand radical is recognized by a character recognizing processor identifying exclusively the left-hand radical and the right-hand radical is recognized by a character recognizing processor 6 identifying exclusively the right-hand radical respectively and each result is fed to a buffer 24. A partial result deciding processor 9 combines the result of analysis and outputs the number corresponding to the character of ''ASE'' to a buffer 26 together with the character number.

Description

Detailed Description of the Invention (1) Technical Field of the Invention The present invention relates to the automatic recognition of handwritten characters (characters with complex patterns such as kanji), and is concerned with the automatic recognition of handwritten characters, etc. (characters with complex patterns such as kanji). This article concerns methods for efficiently carrying out the process.

(2) Conventional technology and problems Conventional character recognition methods include a method in which the target character is segmented, and then radicals such as ``biased'' and ``labor'' are verified separately to extract and judge the characteristics of the character. There is a method that extracts the characteristics of a character from the entire pattern, but the former has the disadvantage of requiring a large amount of processing time because it analyzes each segment, while the latter has the disadvantage that it takes a long time to process, while the latter has common problems such as the same bias and effort. The drawback was that the rate of misrecognition between characters with radicals (for example, the character ``sweat'' and the character ``dirt'') was high.

(3) Purpose of the Invention In view of the above-mentioned drawbacks of the conventional art, an object of the present invention is to provide a character recognition method that has a low rate of misrecognition between characters having a common radical and is capable of high-speed processing. The purpose is

(4) Structure and object of the invention According to the present invention, as described in the claims, a plurality of systems consisting of processors each recognizing a different character or a character in a different character group are provided. Connected in parallel, the same character is input from the character string to be processed into each of the plurality of systems at the same time, each unit performs U recognition processing in parallel, and the processor that has finished its own processing immediately This is achieved by a character recognition method characterized in that it starts possible processing relating to the recognition of the next unprocessed character in the character string.

(5) Embodiment of the Invention FIG. 1 is a block diagram of an embodiment of the present invention, in which 1 is an observation character generation processor, 2 to 8 are character recognition processors, 9dO is a partial result determination processor, and 11 is a result determination processor. The processor, 12 to 29 are buffers, and 30.31 is a bus.

In FIG. 1, an observed character generation processor 1 scans a target character to generate observed character information, and attaches a character number (a number that identifies the observed character) for the character. The system that includes the character recognition processor 2.5.6 specializes in recognizing kanji that can be divided into ``partial'' and ``labor'', and the system that includes the character recognition processor 3.7.8. The system that includes the character recognition processor 4 specializes in recognizing kanji that can be divided into crowns and legs.
The system including the above is set to recognize characters other than these, and consideration has been given to the optimal algorithm for recognizing the characters in charge of each character. Further, the result determination processor 11 ultimately determines the recognition result for the character based on the information obtained from each group.

To explain the operation in the block diagram of FIG. 1 using the character ``6'' as an example, the observed character generation processor 1 observes the character ``6'', performs the preprocessing described above, and buffers the character number and observed character information. 12 and bus 30 to buffers 13-15, respectively.

The character recognition processor 2 receives the observed character information (
``Sweat'') is segmented and divided into character information for bias and labor.These are assigned character numbers and are processed by the character recognition processor 5, which specifically identifies bias for bias, and character recognition for specialized for labor and emperor. Each result is recognized by the processor 6 and each result (number for the type of bias and number for the type of effort) is sent to the buffer 24 via the buffer 20.21.The partial result determination processor 9 processes these analysis results. The combined number corresponding to the character "sweat" is output to the buffer 26 together with the character number as a result of the analysis.

On the other hand, the character recognition processor 3 attempts to segment the given observed character information (``sweat'') and extract the crown part from it, but since it cannot do so, it notifies the partial result determination processor 10 of this fact. Then, the partial result determination processor 10 outputs the character number and information indicating that the analysis has failed to the buffer 27.

Furthermore, the character recognition processor 4 is unable to analyze the given observed character information ("sweat #") and outputs the character number and information indicating that the analysis has failed to the buffer 28.

The result judgment processor 11 uses these buffers 26 to 28.
The character recognition process for the character is thereby terminated. During this time, for example, the character recognition process for the character is terminated. ” is input, the observed character generation processor 1 performs preprocessing as described above, and when the character number and observed character information of the character “u” are sent to the buffers 13 to 15, the character recognition processor If 3 is empty (the process in charge has already ended by failing to extract the crown from the character "sweat" and reporting that fact), there is still recognition about "sweat" in other systems. 1 immediately regardless of whether processing continues.
The character "U" is segmented and divided into a crown and legs, and the character recognition processor 7 starts recognition processing for the crown, and the character recognition processor 8 starts recognition processing for the legs.

In this embodiment, as a character recognition processor system, a system for recognizing characters that can be divided into unilateral and hard characters, a system for recognizing characters that can be divided into crowns and legs, and a system for recognizing other characters will be described. In addition, there are systems that recognize only specific characters, systems that specialize in recognizing specific characters, such as karate, crowns, legs, etc., and systems that target only patterns that include specific conditions. Various combinations of systems can be considered.

In either case, by building an algorithm that can most efficiently recognize characters that meet the specific conditions that the system needs to recognize, it is possible to create a character recognition device that is fast and has a high recognition rate. be.

(6) Effects of the Invention As explained in detail above, according to the method of the present invention, only a relatively narrow range of characters, such as individual characters or characters with certain conditions, can be recognized. Since the algorithm can be set for this purpose, there is an advantage that a device with a high recognition rate can be obtained. In addition, by decentralizing processing, the algorithm of each group becomes simpler, and further parallel processing can be performed within each system (for example, processing bias and labor in parallel). The process has already finished.
The next input character can be pre-processed using the idle processor, so high-speed character recognition can be performed, which is very effective.

[Brief explanation of drawings]

FIG. 1 is a block diagram of one embodiment of the present invention.

Claims (1)

[Claims] A plurality of systems each consisting of a processor that recognizes a different character or a character in a different character group is connected in parallel, and each of the plurality of systems simultaneously receives the same characters as the string to be processed. Characters that are input and each group performs recognition processing in parallel, and the processor that finishes its assigned processing immediately starts possible processing related to recognition of the next unprocessed character in the character string. Recognition method.