JP2010072734A - Handwritten character recognition device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a handwritten character recognition device recognizing a handwritten character with writing time information obtained by a digitizer or the like with sufficiently high recognition rate. <P>SOLUTION: The device includes: a stroke acquisition part 1 which acquires a writing stroke from coordinate information input to a coordinate input device; a template storage part 2 in which stroke templates are registered; a stroke normalization part 3 which normalizes the writing stroke; a coordinate conversion temporary storage part 4 which stores the content of coordinate conversion in the normalization; a template determination part 5 which compares the normalized writing stroke with the templates selects a corresponding template; a template restoration part 6 which restores the template in a size and position according to the writing stroke by performing inverse conversion of the coordinate conversion in the normalization thereto; a stroke imaging part 7 which images a template group composed of the restored templates; and a transmission part 8 which transmits the stroke image to an OCR device. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a character recognition device that is acquired in real time by a device such as a digitizer and recognizes writing.

Conventionally, many optical character recognition (OCR) devices were originally used as postal code recognition devices such as those disclosed in Patent Document 1 and the like. However, as digital cameras and scanners are widely used, they become common and low. The price is supplied. However, since the writing time information represented by the order of writing cannot be used, the recognition rate is low, and the recognition of handwritten characters is not practical, especially 50% or less.
On the other hand, writing data acquired with a digitizer can use writing time information, so the character recognition rate may be higher than an OCR device that recognizes characters from image information acquired with a scanner. Only a small amount is produced because the equipment is necessary and the generality is low and the number of users is small. Therefore, the price is also expensive.
Conventionally, there has been a character recognition device that images a writing locus obtained as time-series coordinate information from a digitizer and gives it to an inexpensive OCR device. However, since the writing time information cannot be used with this device, only a recognition rate as low as that of the conventional OCR device can be obtained, which is not practical.
JP 09-081664 A

  The present invention solves the above problems and provides a handwritten character recognition device capable of recognizing handwritten characters with writing time information obtained from a digitizer or the like with a sufficiently high recognition rate while using an inexpensive OCR device. Objective.

In order to solve the above problems, the character recognition device of the present invention includes a stroke acquisition unit that acquires a writing stroke from coordinate information input to the coordinate input device, and a plurality of writing strokes. A template storage unit registered as a template; a stroke normalization unit that normalizes the writing stroke acquired by the stroke acquisition unit by performing coordinate conversion; and a coordinate conversion temporary storage unit that stores the processing content of the coordinate conversion; A template determination unit that selects a template corresponding to the writing stroke based on the similarity obtained by comparing the normalized writing stroke with the template stored in the template storage unit; and On the other hand, the coordinate transformation stored in the coordinate transformation temporary storage unit is inversely transformed, and the writing stroke is A template restoration unit that restores the size and position according to the template, a stroke imaging unit that images a template group constituted by the restored templates, and a communication unit that provides the imaged stroke image to the OCR device And a handwritten character recognition device.
The invention according to claim 2 is the handwritten character recognition apparatus according to claim 1, wherein the template discriminating unit is normalized when the normalized writing stroke and the template are superimposed. The template having the smallest area area to be enclosed with the writing stroke is selected.

  The inputted writing stroke is converted into a stroke registered in advance and easy to perform OCR processing, and then imaged. For this reason, the OCR device can process the handwritten characters that have been clarified, and a sufficient recognition rate can be obtained even with an inexpensive OCR device.

Next, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram of a character recognition apparatus showing an embodiment of the present invention.
The character recognition device of the present invention includes a writing stroke acquisition unit 1 that acquires a writing stroke from coordinate information input to a coordinate input device 12 such as a digitizer, and a stroke template storage unit 2 in which a plurality of writing strokes are registered in advance as templates. The coordinate conversion process is performed on the input writing stroke, the size normalization unit 3 for normalizing the size and length, and the magnification at the time of the coordinate conversion process performed by the writing stroke normalization unit 3 are stored. The coordinate conversion temporary storage unit 4 compares the normalized writing stroke with the registered stroke template (pattern matching), and selects a stroke template corresponding to the acquired writing stroke based on the similarity. 5. For the selected stroke template, A stroke template restoration unit 6 that performs a reverse transformation of the coordinate transformation stored in the transformation temporary storage unit 4 and restores the size and position according to the writing stroke, and a restored stroke template group constituted by the restored stroke template. A stroke imaging unit 7 for imaging, and a stroke image communication unit 8 for supplying the imaged stroke image to an external OCR device 13 are configured.
These components are connected to each other by a data bus 10 and exchange necessary data as appropriate. The writing stroke acquisition unit 1 is connected to the data bus 10 via the device interface 11.
Subsequently, processing in each component will be described in more detail.
The writing stroke acquisition unit 1 samples, for example, coordinate information detected by the digitizer 12 at regular intervals such as 100 Hz, and includes a coordinate position indicated by an X coordinate and a Y coordinate, a time when writing is performed, and the like. Get information.

なお、デジタイザとは、位置を指示するためのペン型の装置と、ペンの検出するための板状の装置を組み合わせた装置であり、表１において、ペンＵ／Ｄの項目は、ペンの板状装置に対するＵｐ／Ｄｏｗｎを示す。
Ｄｏｗｎの状態からＵｐの状態にかわるまでの座標情報によって、１つの筆記ストロークが構成されるものである。 An example of the acquired stroke information is shown in Table 1 below. As shown in Table 1, the writing stroke is acquired in time series in the order of sampling.
[Table 1]

The digitizer is a combination of a pen-type device for indicating a position and a plate-like device for detecting a pen. In Table 1, the item of pen U / D is a pen plate. Shows Up / Down for the device.
One writing stroke is constituted by coordinate information from the Down state to the Up state.

FIG. 2 is a diagram illustrating a typical example of a stroke template registered in the stroke template storage unit 2.
In the stroke template storage unit 2, the stroke template is represented by the X coordinate and the Y coordinate as with the writing stroke. In the stroke template, a plurality of typical strokes are registered so that character recognition is easy.
As shown in FIG. 2, in the stroke template storage unit 2, for example, a plurality of stroke parts (templates) that constitute characters such as (a) straight line, (b) L-shape, and (c) S-shape are registered. . In addition, the size or length of these stroke templates is normalized and registered in advance so as to be within a certain range so that (comparison) pattern matching processing with subsequent writing strokes is easy. ing.
Note that the stroke template is a stable stroke suitable for OCR, such as that obtained in type and clear text.

The writing stroke normalization unit 3 receives the writing stroke acquired by the writing stroke acquisition unit 1, and normalizes the length and size so that pattern matching processing with a later stroke template is easy.
The normalization process can be easily performed by using a known means.
At this time, the coordinate conversion temporary storage unit 4 temporarily holds the magnification in the coordinate conversion accompanying the normalization process and the start point coordinates of the writing stroke.
The stroke template determination unit 5 compares (patterns match) the writing strokes normalized by the writing stroke normalization unit 3 with a plurality of stroke templates registered in the stroke template storage unit 2, and inputs the writing strokes Select the stroke template that most closely resembles (highly similar).

FIG. 3 is a diagram illustrating an example of pattern matching.
When there is a normalized writing stroke and stroke templates (a) to (c) inputted from a digitizer or the like, considering the distance between the writing stroke and the stroke template, they are surrounded by both when they are superimposed. The similarity can be expressed by the area of the region.
In this case, it can be said that the stroke template (c) having the smallest area area surrounded by both has the highest similarity, and the stroke template (c) is selected as the template corresponding to the writing stroke.
In the normalization by the writing stroke normalization unit 3, the stroke template restoration unit 6 performs, for example, coordinate conversion for multiplying the size of the writing stroke by N, and the value of the magnification is the coordinate conversion temporary storage unit 4. Is stored in the selected stroke template, the selected stroke template is multiplied by 1 / N, and the starting point is moved to the same position as the starting point of the writing stroke.
In other words, the selected stroke template is sent to the stroke template restoration unit 6 and subjected to the inverse transformation of the coordinate transformation stored in the coordinate transformation temporary storage unit 4, and at the same position as the first input stroke, the same size. Now restored.
Next, the restored stroke template is sent to the stroke imaging unit 7 and converted into an image that can be processed by the OCR apparatus.
The stroke imaging unit 7 can use a known process for the imaging process, and configures a stroke image according to the coordinate sequence described by the restored stroke template group. The image data can be selected as appropriate, such as a known bitmap image or TIFF image.
The normalized written writing stroke image is sent to the OCR apparatus and used as data that can be subjected to the conventional OCR processing.

FIG. 4 is a diagram showing an example of the flow until text information is recognized in the OCR device based on the writing stroke acquired in the present invention.
(A) Based on the acquired writing stroke, (b) a corresponding stroke template is determined, and this stroke template is imaged and sent to the OCR device 13 for recognition processing to be identified as text.
As described above, according to the character recognition device shown in the present embodiment, a sufficiently high recognition rate can be obtained even by using an existing inexpensive OCR device by converting handwritten stroke information into an appropriate stroke and imaging it. An apparatus for recognizing handwritten characters can be provided.

FIG. 5 is a flowchart for explaining processing in the handwritten character recognition apparatus of the present invention.
First, the writing stroke acquisition part 1 acquires a writing stroke from the coordinate information input into the digitizer etc. (S101). Next, the stroke normalization unit 3 performs coordinate conversion processing on the input writing stroke, normalizes the size and length (S102), and writes the stroke conversion normalization unit 3 to the coordinate conversion temporary storage unit 4. The coordinate conversion process performed by (1) is stored (S103).
Next, the stroke template discriminating unit 5 compares (pattern matching) the plurality of strokes registered in advance in the stroke template storage unit 2 with the normalized writing stroke and the registered template (S104), and the similarity The stroke template corresponding to the writing stroke is selected from the degree (S105).
Further, the stroke template restoration unit 6 performs reverse transformation of coordinate transformation stored in the coordinate transformation temporary storage device on the determined stroke template, and restores the size and position according to the writing stroke (S106).
Then, the stroke imaging unit 7 images the restored stroke template group constituted by the restored stroke template (S107).
Finally, the stroke image communication unit 8 gives the imaged stroke image to the external OCR device (S108).

The block diagram of the character recognition apparatus which shows the Example of this invention. The figure which shows the typical example of the stroke template registered. The figure which shows the example of pattern matching. The figure which shows an example of a flow until text information is recognized in an OCR apparatus based on the writing stroke acquired in this invention. The flowchart explaining the process in the handwritten character recognition apparatus of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Written stroke acquisition part, 2 stroke template memory | storage part, 3 stroke normalization part, 4 coordinate transformation temporary storage part, 5 stroke template discrimination | determination part, 6 stroke template restoration part, 7 stroke imaging part, 8 stroke image communication part, 10 Data bus, 11 Device interface

Claims (2)

A stroke acquisition unit that acquires a writing stroke from the coordinate information input to the coordinate input device;
A template storage unit in which a plurality of writing strokes are registered as templates;
A stroke normalization unit that normalizes by performing coordinate transformation on the writing stroke acquired by the stroke acquisition unit;
A coordinate transformation temporary storage unit for storing the processing content of the coordinate transformation;
A template determination unit that selects a template corresponding to the writing stroke based on the similarity obtained by comparing the normalized writing stroke and the template stored in the template storage unit;
A template restoration unit that performs reverse transformation of coordinate transformation stored in the coordinate transformation temporary storage unit for the selected template, and restores the size and position according to the writing stroke;
A stroke imaging unit for imaging a template group constituted by the restored templates;
A communication unit for providing an imaged stroke image to the OCR device;
The handwritten character recognition apparatus characterized by consisting of. The handwritten character recognition device according to claim 1,
The said template discrimination | determination part selects the template with the smallest area area enclosed with the said normalized writing stroke, when the said normalized writing stroke and the said template are overlapped, The handwritten character recognition apparatus characterized by the above-mentioned.

Images