JPH063613B2 - Handwriting recognition device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a handwritten character recognition device capable of easily and highly accurately recognizing a character pattern handwritten and input via a coordinate input device such as a tablet.

[Technical background of the invention and its problems]

Recently, a handwritten character recognition device has been developed which recognizes a character pattern input by handwriting through a coordinate input device such as a tablet, using the stroke information.

As shown in FIG. 6, this handwritten character recognition device guides information of a character pattern written and input from a character input unit 1 composed of a coordinate input device to a pre-processing unit 2 to remove input noise and reduce the size of characters. After performing preprocessing such as normalization, the feature extraction unit 3 detects the feature of the stroke forming the input character pattern. This stroke feature extraction is performed, for example, by obtaining the position coordinates of the start point, middle point, and end point of each stroke.

Then, the recognition processing unit 4 inputs the stroke information of the input character pattern whose features have been extracted, rearranges the stroke order according to a predetermined stroke order rule, and then recognizes the recognition target category registered in the standard pattern dictionary 5. The information on the strokes forming each of the standard patterns is compared with the information on the strokes of the input character pattern.

To match the standard pattern with the input character pattern, for example, first, a standard pattern having the same number of strokes as the stroke number of the input character pattern is selected. Then, between the selected standard pattern and the input character pattern, the feature similarity or distance between corresponding strokes is calculated according to the stroke order.

The collation result of the input character pattern thus collated with each of the standard patterns is given to the identification unit 6.
The identification unit 6 obtains, for example, the standard pattern having the largest similarity value or the standard pattern having the shortest distance value as the recognition result for the input character pattern.

The input character pattern is given as a series of strokes written on the tablet (character input unit) 1 as shown in FIG. 7, for example. Therefore, some authors write characters in a stroke order different from the regular stroke order. Therefore, as described above, the writing stroke order is rearranged so that the input pattern and the standard pattern are matched.

However, in handwritten character recognition, there is not only the problem of the stroke order but also the problem of the direction of the writing stroke. For example, the first of the character pattern "hi" shown in FIG.
When it comes to strokes, the stroke is normally written to the right, but some people write it to the left.

Therefore, in the prior art, when associating a dictionary stroke and a writing stroke, for example, as shown in FIG. 8, a dictionary in which the direction of the input stroke is reversed is prepared. Then, both the writing stroke and the stroke in which the direction is reversed are collated with the dictionary stroke, and a stroke having a small difference is selected as the stroke in the original direction.

By doing so, it is possible to correct the error in the stroke direction and to effectively recognize the written and input character pattern. However, for that purpose, the collation process for the input stroke and the process for the reverse stroke are doubly required, which requires twice the amount of calculation as compared with the conventional character recognition that does not consider the direction of the stroke. There was a problem such as doing.

[Object of the Invention]

The present invention has been made in consideration of such circumstances, and an object thereof is to simplify a character pattern handwritten and input through a coordinate input device such as a tablet without being concerned with an error in stroke direction. Another object of the present invention is to provide a handwritten character recognition device that can recognize with high accuracy.

[Outline of Invention]

The present invention detects strokes of a character pattern input by handwriting via a coordinate input device such as a tablet, and collates the information of each stroke with the stroke information of the recognition target category registered in the standard pattern dictionary to write the strokes. In the handwritten character recognition device that recognizes the input character pattern, the stroke information that may be input with the stroke direction forming the standard pattern of the recognition target category reversed is registered in the standard pattern dictionary. The stroke information may be used to detect strokes that may be input with the orientation reversed in the strokes of the handwritten character pattern, and the orientation may be input with the orientation reversed. The stroke writing direction (stroke direction) is set to the standard pattern It is adapted to be used for collation with the standard pattern dictionary after matching the direction of the peak.

〔The invention's effect〕

Thus, according to the present invention, even if there is an error in the direction of the writing stroke, it can be used for matching with the standard pattern dictionary after being matched with the direction of the stroke registered in the standard pattern dictionary. Does not increase the calculation amount of. Moreover, since the process for correcting the stroke direction is performed only for the stroke in which the stroke direction may be reversed, the process can be performed easily and efficiently.

Therefore, a character pattern input by handwriting can be simply and efficiently
Practical great effects such as high-precision recognition can be obtained.

Example of Invention

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a main part of the apparatus of the embodiment, in which the same parts as those of the conventional apparatus shown in FIG.

The device of the present embodiment is characterized in that the standard pattern dictionary 5 is provided with the feature information of the strokes forming the standard pattern as the feature pattern dictionary 11 as in the conventional device, and is also provided with the stroke order dictionary 12 for the standard pattern. In addition to this, the stroke inversion dictionary 13 is provided.

The recognition processing unit 4 is characterized by including a stroke reversal processing unit 16 in addition to the stroke order rearrangement processing unit 14 and the matching processing unit 15.

The stroke order rearrangement processing unit 14 is the stroke order dictionary.
The order of input strokes is rearranged according to the stroke order rule stored in 12. Also, the matching processing unit 15 collates the stroke feature information that forms each standard pattern of the recognition target category registered in the feature pattern dictionary 11 with the stroke feature of the input character pattern, respectively, and obtains the input character pattern. Is to calculate the similarity or distance between the standard pattern and the standard pattern. In the conventional device, the information of the input character pattern processed by the stroke order rearrangement processing unit 14 was supplied to the matching processing unit 15, but in this device, the input characters processed by the stroke order rearrangement processing unit 14 are input. After the pattern inversion processing section 16 processes the pattern information, the matching processing section 15 collates it with the standard pattern.

Now, the stroke of the character pattern which is written and input through the coordinate input device 1 such as a tablet is, for example, as shown in FIG.
As shown in (d), it can be roughly classified into four types according to its orientation. That is, it can be roughly classified into a stroke in the upward right direction, a stroke in the downward left direction, a stroke in the downward right direction, and a stroke in the upward right direction.

Therefore, when the direction of each of these strokes is decomposed into a horizontal direction X and a vertical direction Y, information on the direction is (X; ±,
It can be expressed as Y; ±). However (+)
Indicates right or upper, and (−) indicates left or lower.

However, using information on the direction of such strokes,
For example, as shown in FIG. 3, the writing directions of the first to third strokes that form the character pattern “H” are expressed as follows.

First stroke ... (X;-, Y ;-) Second stroke ... (X; +, Y; +) Third stroke ... (X;-, Y ;-) Here, for example, in the above three strokes, When there is a possibility that the stroke of 1 is written with its direction reversed, the stroke inversion dictionary 13 of the standard pattern dictionary 4
For example, information "X-" is stored as stroke reversal information.

In the above "X-", the direction of the first stroke is X.
It is reversed in the (horizontal) direction, that is, the first stroke is originally written rightward, but it may be written leftward.

The standard pattern dictionary 4 is configured as shown in FIG. 4, including such stroke reversal information. That is, the standard pattern dictionary 4 is configured by registering the character code A indicating the recognition category, stroke inversion information B, and stroke characteristic information C for each standard pattern.

The stroke reversal information B is given as information on the reversal direction of each stroke forming a standard pattern, and when the strokes may be reversely input by writing, the stroke reversal information B is also written in the opposite direction. For a stroke that has no possibility, it is given as information "NN", for example.

The stroke characteristic information C is given as data indicating the position coordinates of the start point, the middle point, and the end point of each stroke.

The stroke reversal processing unit 16 described above uses such stroke reversal information to first extract strokes that may be written in the opposite direction from the strokes that form the input character pattern, and then extract the strokes. About the information of the stroke direction written and input. When the direction of the stroke is the same as the direction of the stroke in the standard pattern, the feature information about the input stroke is output as it is. When the direction of the stroke in the standard pattern is opposite, for example, the feature information of the input stroke is used. The position coordinate data of a certain start point is exchanged with the position coordinate data of the end point, the direction of the stroke is reversed, and the output is made in accordance with the direction of the stroke of the standard pattern.

That is, in the stroke reversal process, for example, when a character pattern "H" is written and input as shown in Fig. 5, first, the standard pattern dictionary 5 of the recognition target category "H" shown in Fig. 4 is referred to. However, it is performed after the possibility that the writing direction of the first stroke may be reversed according to the stroke inversion information. Then, with reference to the information on the direction of the first stroke, it is required that the direction of the first stroke that has been written and written is opposite to the standard stroke. As a result, the position coordinate information of the start point and the end point of the feature information of the stroke are exchanged, and the orientation of the stroke is matched with the orientation of the stroke in the standard pattern.

The matching processing unit 15 performs the matching process between the stroke information of the input pattern thus stroke-reversed and the stroke information of the standard pattern registered in the standard pattern dictionary 5. This matching process is performed by calculating the similarity or distance between each corresponding stroke.

In this way, according to the present device, for a stroke in which the stroke direction may be reversed, the matching process with the standard pattern is performed after the stroke direction is matched with the stroke direction of the standard pattern. The direction of the input stroke that is matched with the standard pattern is always aligned with the direction of the standard stroke. Therefore, unlike the conventional case, it is not necessary to perform the matching process for each of the writing stroke and the stroke of which the direction is reversed, so that the calculation amount does not increase. Moreover, the direction of the stroke is adjusted only for the stroke in which the writing direction may be reversed. Therefore, for example, there is no possibility that "so" and "n" will be erroneously recognized.

That is, when stroke information in which the stroke direction of the input character pattern "SO" is reversed and matching with a standard pattern is performed, "SO" is erroneously recognized as "N" depending on the characteristic information of the stroke. There is a possibility that it will be done. However, in this device, since there is no possibility that the writing stroke of "So" is written in the opposite direction, the reversal of the writing stroke is not permitted in the end, and the error as described above is not caused. It is possible to obtain an effect such as no recognition.

Therefore, the character pattern input by handwriting can be simply and efficiently
It is possible to recognize with high accuracy.

The present invention is not limited to the above embodiment. For example, in a device in which the stroke order is limited, the stroke order rearranging process can be omitted. In addition, the stroke information may be given only for strokes that may have opposite stroke directions. Other,
The present invention can be variously modified and implemented without departing from the scope of the invention.

[Brief description of drawings]

FIG. 1 is a schematic view of the essential parts of an apparatus according to an embodiment of the present invention.
FIG. 4 is a diagram showing the direction of a stroke input by handwriting, FIG. 3 is a diagram showing information on the stroke direction of the character pattern “H”, and FIG. 4 is a diagram showing a configuration example of a standard pattern dictionary.
FIG. 6 is a diagram showing stroke direction information of a character pattern “H” input by handwriting. FIG. 6 is a diagram showing a schematic configuration of a conventional device. FIG. 7 is an example of handwriting input of a character pattern on a tablet. FIGS. 8A and 8B are views for explaining the process for the stroke direction in the conventional device. 1 ... Character input device (coordinate input device), 2 ... Preprocessing unit, 3
... feature extraction unit, 4 ... recognition processing unit, 5 ... standard pattern dictionary, 6 ... identification unit, 11 ... feature pattern dictionary, 12 ... stroke order dictionary, 13 ... stroke inversion dictionary, 14 ... stroke order rearrangement processing unit, 15 ... matching Processing unit, 16 ... Stroke reversal processing unit.

Claims (3)

[Claims] 1. A stroke of a character pattern input by handwriting is detected, and information of each stroke is collated with stroke information of a recognition target category registered in a standard pattern dictionary to recognize the character pattern input by handwriting. In the handwritten character recognition device, the stroke information that may be input by reversing the direction of the stroke forming the standard pattern of the recognition target category is registered in the standard pattern dictionary, and the stroke information is registered. A means for detecting a stroke that may be input by reversing the direction in the stroke of the handwritten character pattern by using the stroke, and a direction that is detected by this means may be reversed and input. And means for adjusting the stroke direction to the stroke direction of the standard pattern using the stroke information. A handwritten character recognition device characterized by the above. 2. The stroke information of the recognition target category registered in the standard pattern dictionary is composed of information indicating the direction of the stroke forming the standard pattern of the recognition target category and information indicating the positional coordinate characteristics of the stroke. The handwritten character recognition device according to claim 1. 3. The collation between a handwritten input character pattern and a standard pattern is performed by calculating the similarity or distance between corresponding strokes.
Handwriting character recognition device described in the paragraph.