JPS601676B2 - character recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides four types of characters, such as dakuten, (〃), semi-dakuten (o), and period (.

), etc. The present invention relates to a character recognition device suitable for recognizing characters such as Small-sized characters, dakuten (''), half-dakuten (o), period (.

) etc. are written in a deformed manner, there are few features that are effective for character recognition, so it is difficult to determine a highly accurate answer based on the features of that character alone. However, in this case, it is possible to narrow down the answer candidates to two or three. Further, such small-sized characters often have some kind of correlation (relative position of characters, connection of characters) with the characters on the left side. For example, dakuten and handakuten are generally written in the upper half of the character to the left, and periods are usually written in the lower half, and the characters to the left of the dakuten are written in ikigyo and sagyo. It is one of the characters T, H, and U, and the character to the left of the handakuten is usually not any other character. The method of the present invention is a recognition method that uses this relative information with the adjacent character on the left as a feature.

The present invention will be explained below with reference to the figures.

The figure shows an embodiment of the present invention, in which 1 is a character buffer that stores signals from a CCD image sensor, 2 is the height of a character,
a position detector for determining the width and base point; 3 a memory for storing information from the position detection section; 4 a feature extraction section for extracting contour features of characters; 5 a memory for storing contour features extracted by the feature extraction section; 6 is a determination unit that compares the pattern in the dictionary 7 with the contents of the memory 5; 8 is a control unit; 9 is a memory that stores character position information; An editing section 11 determines the answer, a memory 11 stores the pattern in the dictionary selected by the determining section 6, and a memory 12 stores the answer.

To explain the operation, character information photoelectrically converted by the CCD image sensor is stored in the character buffer 1. The character information in the character buffer 1 is input to the position detector 2 and the feature extraction unit 4, and the height, width, and base point position of the character are determined by the position detector 2, and these data are stored in the memories 3 and 9.
is input.

The data in the memory 3 is further input to a feature extraction section 4, and based on this data, the contour features of the characters input from the character buffer 1 are extracted.

On the other hand, the data stored in the memory 9 are also used for small-sized character recognition, which will be described later.

The outline of the character extracted by the feature extraction unit 4 is stored in the memory 5, and compared with the pattern in the dictionary in the determination unit 6. As a result of the comparison, the selected pattern is stored in the memory 11.
The information is input to the editing section 10 via '. If a relative information check instruction is not added to the selected pattern, the pattern is input to the memory 12 from the editing section 10 as is. Here, relative information check instructions are added to the patterns in the dictionary for the 4-shaped characters as described above. If this check instruction has been added, the editing department 10 uses the character position information stored in the memory 9 and the connection with the adjacent character on the left stored in the memory 2 to create the final answer. Output. In other words, when three characters are listed as answer candidates, as shown below: 1st candidate: 2nd candidate: half-voiced mark: 3rd candidate: period. Compare and determine whether it is a dakuten, a half-dakuten, or a full stop.

If it is in the upper half of the character to the left, it is a dakuten or handakuten, and if it is in the lower half, it is a period. If there is no character to the left, it is impossible to determine and the character is clicked to be ejected. Once the answer candidates have been narrowed down to either voiced or handakuten, the next step is to examine the connection between the characters on the left. As a result, if the character on the left is a character in the C line, there are both possibilities, so the candidate with the highest priority (in this case, the handakuten) is selected as the answer. Also, if the character to the left is either ikigyo, sagyo, yugyo, or u, it will be a voiced mark. If it is neither of the above, it is impossible due to the connection of characters, so it will be rejected. The above operation control is performed by the control section 8. As described above, according to the present invention, since a small-sized character is determined based on its position and the connection with the adjacent character on the left, it is possible to accurately recognize the character.

A simple explanatory diagram of the data drawing shows an embodiment of the present invention, in which 1 is a character buffer, 2 is a position detector, 3, 5, 9, and 11 are memories, 4 is a feature extraction section, 6 is a determination section, and 7 is a Dictionary, Ichikawama Editorial Department.

Claims (1)

[Claims] 1. A position detection means for detecting the position of a read character, an extraction means for extracting the characteristics of the character, and a determination section for selecting similar or similar patterns from patterns in the dictionary based on the characteristics from the extraction means. and an editing section that outputs a final answer based on the relative position with the adjacent character on the left or the connection with the adjacent character on the left when a relative information check instruction is added to the output from the determining section. A character recognition device featuring: