JPS5955584A - Time reducing system for recognizing process - Google Patents

Abstract

PURPOSE:To improve the working efficiency, by sorting the input characters in terms of the storke regarded effective and selected out of a recognizing standard pattern and then recognizing the feature quantity of all strokes with obtained candidate categories. CONSTITUTION:A preprocess 1 is given to the input character for elimination of noises, smoothing, etc., and the stroke number S is fed to a standard pattern memory 6 and a candidate stroke storage device 7. The memory 6 sends the stored features to a feature extracting device 2 and an inter-stroke distance calculating device 3 in order of stroke numbes designated to the device 7 in terms of the first category of a standard pattern of S strokes. The device 2 extracts the features of only the stroke fed from the memory 6 and sends them to the device 3. Then the minimum stroke is decided 4 from the result of calculation and sent to a candidate deciding device 5, and at the same time the same processing is carried out also with the remaining strokes of a character under collation. The result of this processing is stored in the device 5. Other standard patterns are also processed in the same way, and the recongnition is decided by the pattern distance.

Description

Detailed Description of the Invention (1) Description of the field to which the invention pertains The present invention describes nine characters as a set of strokes, and describes nine characters as a set of strokes. The present invention relates to a character recognition device that recognizes input characters by calculating distances corresponding to strokes.

(2) Description of Prior Art In conventional devices of this type, a method has been used in which a classification feature is prepared separately from the features used in recognition processing for classification processing to narrow down candidate categories. Therefore, in order to perform two-classification processing, a storage device is required to store the classification features, and the procedure for extracting the classification percentage features from the input characters separately from the recognition features increases, which requires a device configuration. The drawback was that it was complicated.

(3) Purpose of the Invention In order to eliminate these drawbacks, the present invention specifies a group of strokes considered to be effective from among the strokes of a standard pattern used for recognition, and uses only feature amounts related to the specified strokes. First, input characters are classified, and then the obtained candidate categories are recognized using feature amounts related to all strokes.

below.

The drawings will be explained.

(4) Description of structure and operation of the invention FIG. 1 shows an embodiment of the invention, in which 1 is a character input preprocessing device, 2 is a feature extraction device, 3 is an inter-stroke distance calculation device, and 4 is a minimum stroke determination device. Device.

5 is an inter-pattern distance and candidate determination device, 6 is a standard pattern memory, and 7 is a candidate stroke storage device.

To operate this, first, for each standard pattern registered in the standard pattern memory 6, sort the strokes in order of importance (for example, if the stroke length is the importance, the stroke length The stroke numbers are registered in the candidate stroke storage device 7 in descending order of the number of strokes. This can be provided externally by signal 106.

Next, operations for characters input into the device 1 will be explained in order. Noise is removed from the entered characters.

Preprocessing is performed using smoothing, size normalization, and known techniques. The number of strokes (S and Suru) of the input character is input into the standard pattern memory 6 and the candidate stroke storage device 7 by the stroke number notification signal 103.

The device 6 selects stroke number 1 specified by the device 7 for the first category of standard patterns belonging to 8 strokes.
1, information regarding the description of stroke characteristics stored in the device 6 (for example, Japanese Patent Application No. 53-153999)
According to the number of feature points that approximate and represent the stroke)
is sent to the feature extraction device 2. At the same time, the feature amount of the stroke (for example, according to Japanese Patent Application No. 51-083733, the (x, l/) coordinate values of the feature point of the stroke and the value of its variance) is sent to the inter-stroke distance calculation device 3. The feature extraction device 2 extracts features from each stroke making up the input character according to the information regarding the description of stroke features from the device 6. The extracted features of each stroke of the input character are sent to the inter-stroke distance calculation device 3.

The inter-stroke distance calculation device 3 uses the stroke feature amount from the device 6 to calculate the inter-stroke distance between each stroke of the input character and the first read stroke of the standard pattern.

Note that the distance between strokes can be calculated using a known technique. The distance between each of these strokes is dI,
d2+..., d8. (S is the number of strokes of the input character)
Shown as The distances d1 to ds between each stroke are sent to the minimum stroke determination device M4.

The minimum stroke determining device 4 determines the minimum value among d1 to d6 and sends it to the inter-pattern distance and candidate determining device 5. At the same time, the device 4 causes the device 6 to repeatedly execute the above two or more processes for the remaining strokes of the standard pattern currently being compared in the order of candidate strokes - strokes specified in the storage device 7. The signal 104 is sent out. As a result of this repetition, for each stroke of the standard pattern currently being compared, the distance between the strokes of the nearest input character is sequentially determined and stored in the device 5. Note that this iterative process is performed based on the number of strokes used for the classification process (
M) is performed a number of times equal to M).

Between the standard pattern "Jin" and the input character 't', the distance between the strokes and the distance between the strokes obtained when the strokes of the standard pattern are used in stroke order (when the importance of strokes is set as stroke order) in two strokes. The minimum value determined for each stroke of the standard pattern is shown in FIG.

Thereafter, the inter-pattern distance and candidate determining device 5 was determined for each stroke of the standard pattern currently being compared. The distance between the minimum strokes (d in Figure 2)
1 (indicated by z・d2b) and calculates the number of strokes used for the two-classification process M (in Fig. 2, M
=2).

This value is stored as the inter-pattern distance between the standard pattern being matched and the input character (in Figure 2).

p-(d+α+d2b)/2). At the same time, the device 5 sends a signal 105 to the device 6 to start the above-mentioned processes for the next standard pattern. In this way, the inter-pattern distance between the input character and the standard pattern group having the same number of strokes as the input character is determined.

Thereafter, the device 5 selects the number of standard patterns specified by the candidate category number designation signal 109 in descending order of the distance between the obtained patterns as candidate categories that need to undergo recognition processing again, and designates the candidate category name. The device @6 is notified by signal 110.

Note that if the number of candidates is 1 in signal 109, or
07 or signal 108 to use all strokes constituting the standard pattern in the classification process, at this point, the category of the standard pattern that provides the minimum distance between patterns is output as a recognition result, and all strokes are Finish the process. If the signal 109 specifies that two or more candidate categories are to be selected, the device 5 repeatedly performs the same process one or more times using all strokes for each of the candidate categories determined.

(5) Description of Effects As explained in 7 above, according to the present invention, each stroke constituting the standard pattern is sequentially used starting from the stroke that can be said to be effective for recognition. This classification is expected to be effective in recognizing kanji with a large number of strokes, so it has the advantage of significantly shortening processing time.

Also, a dictionary for major classification is often required at all.

There is an advantage that the capacity of the standard pattern memory does not increase, and feature extraction processing for major classification is not required, and the structure of the device is simplified.

Furthermore, in order to extract features from the strokes of input characters based on information about stroke feature descriptions of standard patterns that are stored in advance, a stable identification system can be created for character types that include a mixture of alphabets, numbers, and kanji. Can be configured.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram of one embodiment of the present invention. FIG. 2 is an explanatory diagram of the distance between strokes when two strokes are used. In the figure, 1 is a character input preprocessing device, 2 is a feature extraction device, 3 is an inter-stroke distance calculation device, 4 is a minimum stroke determination device, 5 is an inter-pattern distance and candidate determination device, 6 is a standard pattern memory, and 7-digit candidates It is a stroke memory device. Patent applicant Hiroshi Mori, patent attorney representing Nippon Telegraph Type 1 Public Corporation

Claims (1)

[Claims] Information about each stroke that makes up a character is captured. In a character recognition method that recognizes input characters by calculating the distance between the strokes and information on each stroke of standard patterns registered in a dictionary in advance, it is possible to specify any part or all of the strokes for each standard pattern. means for extracting features of strokes constituting the input character based on feature description information regarding the specified stroke; and means for extracting stroke correspondence between the specified stroke and all strokes of the input character. means for calculating a distance; means for selecting one or more candidate characters that can be a recognition result of the input character; and two characters between the character selected as a candidate and the input character. A recognition processing time reduction processing method characterized by having means for recognizing the input character by calculating distances corresponding to strokes using all the constituent strokes.