JPS61198382A - On-line character input device - Google Patents

Abstract

PURPOSE:To give a high recognizing performance by assembling each modified stroke into a standard pattern so as to correspond the standard pattern stroke to the stroke of an input pattern without excess or deficiency. CONSTITUTION:The deformation of a stroke such as deformation of stroke attended with fluctuation of stroke number, demodulation of stroke continued in erroneous writing order or deformation of stroke shape in one stroke is assembled in the standard pattern together with the stroke number of the standard stroke corresponding to each deformation stroke. The strokes are made corresponded without excess or deficiency so as to correspond the stroke of the standard pattern to the stroke of the input pattern without duplication or excess of the stroke number. The on-line character input device with high recognizing performance independently on the order of writing is obtained by recognizing the character based on the accumulated value of the distance between the strokes in this case.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an online character input device that recognizes characters from handwriting of characters input online.

(Prior art and its problems) Conventionally, online input is possible by allowing variations in the number of strokes (handwriting from when the tablet pen is turned on until it is turned off) and independent of the stroke order (the order in which the strokes are written). There is a method to recognize the written characters. For example, "Transactions of the Institute of Electronics and Communication Engineers" VOJ, J66-D Volume 5 (19
Published on pages 593-600 (May 1983) under the title ``Online character recognition independent of stroke count and stroke order by selective stroke combination'', the following paper describes the following method.

In this method, strokes are correlated between the input handwriting of the next character (referred to as an input pattern) and a standard pattern previously stored in the recognition device. first,
All strokes of the pattern with the smaller number of strokes (referred to as pattern A) between the input pattern and the advancement pattern are respectively associated with the strokes with the smallest distance in the other pattern (referred to as pattern B). Subsequently, the uncorrelated strokes of pattern B are combined with two adjacent strokes in the stroke order that have already been correlated to form a combined stroke. Then, the distance between these combined strokes and the corresponding stroke of pattern A is determined, and the combined stroke with the smaller distance is selected. This selected combined stroke is associated with the stroke of pattern A. In this way, when all the strokes of pattern A and pattern B are associated, the sum of the distances between corresponding strokes is taken as the inter-pattern distance. The standard pattern category with the minimum inter-pattern distance is set as the recognition result.

In this method, variations in the number of strokes are dealt with by combining successive strokes in stroke order. However, in reality, when strokes are continuous, not only are the strokes simply combined, but the shapes of the strokes often change. For example, FIG. 7 is an example of a standard pattern for the character "mouth", and FIG. 8 is an example of a pattern when the number of strokes for the character "mouth" varies. Stroke 82 in FIG. 8 is a continuation of stroke 72 and stroke 73 in FIG. 7, but the shape of the stroke is different from the one that simply combines these two strokes. This method has the disadvantage that patterns with essentially different shapes are compared, and misrecognition is likely to occur.

Furthermore, since strokes are only combined with successive strokes at the beginning, if strokes are written consecutively in an irregular stroke order, misrecognition is likely to occur. ) The object of the present invention is to correct each stroke deformation, such as stroke deformation due to a change in the number of strokes, deformation of nine strokes caused by consecutive strokes in the wrong stroke order, and deformation of the stroke shape within one stroke. By incorporating modified strokes into the standard pattern and making it possible to match the strokes of the standard pattern and the strokes of the input pattern with just the right amount, we have achieved an online character input device that has high recognition performance and does not depend on the next stroke order. be.

(Structure of the Invention) The online character input device according to the present invention includes a handwriting detection section that detects handwriting when writing one character and reads it as an input pattern, and a modified stroke that includes a combination of strokes and a standard stroke pattern of the character to be recognized. The pattern is divided into the stroke number for standard strokes and the stroke number of the corresponding standard stroke (for modified strokes).
(Multiple stroke numbers if it includes a combination of strokes)
Based on the standard pattern memory held as a standard pattern, and the cumulative value of the distance between strokes obtained by associating the strokes of the input pattern and the strokes of the standard pattern so that no duplication or surplus of stroke numbers occurs. It is composed of a recognition section that recognizes characters.

[Principle of the invention]

Here, one principle of the present invention will be explained. The online character input device according to the present invention has a modified stroke pattern as a standard pattern in addition to a standard stroke pattern that is a standard character stroke. Deformed strokes are strokes that have been deformed by K, such as deformation of strokes due to fluctuations in the number of strokes, deformation of nine strokes resulting from successive strokes in the wrong order, and deformation of the shape of strokes within one stroke.

When the strokes of the standard pattern and the input pattern are correctly matched, the character t is recognized based on the inter-pattern pressure at obtained as the cumulative value of the inter-loaf distance. The principle of recognition according to the present invention will be explained below based on an example.

FIG. 2 is an example of a character pattern for "possible". In this example, the characters are made up of standard strokes. However, the "mouth" pattern made up of strokes 22, 23, and 24 in the figure can be modified in three ways, as shown in FIG. 3, for example.

In the figure, stroke 31 is a continuation of strokes 23 and 24, and strokes 32 and 33 are strokes 22 and 33.
゜Edict, 24 consecutive. In the present invention, the standard pattern consists of the standard strokes 21-25 and the modified strokes 31-33.

Figure 4 is an example of the input pattern for the character "possible".
This character pattern consists of strokes 41, 42° 43. The wgs diagram is a table showing the input pattern of the ylK4 diagram, the second factor, and the target of the distance between strokes of the standard pattern shown in FIG. Input pattern horizontally,
The strokes in the vertical direction K [quasi-putter/] are arranged, and the numbers shown as input patterns and standard patterns indicate the strokes in FIGS. 2 to 4. The shape of the stroke is also shown next to it. Stroke numbers 1 to 5 are assigned to the standard strokes of strokes 21 to 25, and stroke numbers of the corresponding standard strokes are assigned to the modified strokes of strokes 31 to 33. If there are multiple standard strokes corresponding to one deformed stroke, 3
, 4, and so on.

As the distance between strokes, a definition as stated in, for example, a patent application filed on January 9, 1989 (hereinafter referred to as cited document (1)) is used. That is, the sum of the Euclidean distances between the coordinate values of the writing points that constitute the strokes is defined as the inter-stroke distance. In addition to this, various definitions of the distance between strokes described in the cited document (1) can be used.

As an example of a deformed stroke, here is a deformation of a stroke due to a change in the number of strokes (stroke 31, 32).
, a series of strokes in the wrong stroke order is shown, resulting in nine stroke variations (stroke o).A modified stroke that involves a variation in the shape of a stroke within one stroke is given the stroke number of the corresponding standard stroke.

Next, based on the inter-stroke distances shown in Figure 5, the stroke that minimizes the cumulative value of the corresponding inter-stroke distance under the condition that the input pattern and the standard pattern strokes are matched without excess or deficiency. Find the correspondence. As a method for efficiently mapping these strokes, the best-first search method (Artificial Intelligence, Iwahi Lecture, Information Science-2
2 (1982), pp. 35-38).

According to this method, the optimal stroke correspondence is searched for with respect to the strokes of the input pattern, with priority given to the correspondence of strokes of the standard pattern with the minimum cumulative value of inter-stroke distances.

At this time, it is determined whether the strokes have been correctly matched or not based on the stroke numbers.

If there are 5 standard strokes as in this example.

Prepare 5-bit compatible memory. Before starting the search, all bits in the corresponding memory are set to O.

Each time a stroke of the input pattern is made to correspond to a stroke of the standard pattern, the bit corresponding to the stroke number in the correspondence memory is set to 1. A stroke with multiple stroke numbers, such as stroke 32, sets multiple corresponding bits to 1. It is assumed that the strokes of the input pattern are not associated with the strokes of the standard pattern that include a stroke number whose bit in the corresponding memory is already 1.

The search ends when all strokes of the input pattern are associated with strokes of any standard pattern, the cumulative value of inter-stroke distances at that time is the minimum, and all bits of the correspondence memory are 1.

The cumulative value of the distance between strokes at this time becomes the distance between patterns.

FIG. 6 is an example of a search tree for the following case in which an optimal stroke correspondence is determined by a best-first search method based on the inter-stroke distances shown in FIG.

The numbers written in the circles representing the nodes are the cumulative values of the distances between strokes. A new node is created by extending the branches of the search tree starting from the node with the minimum cumulative value of the distance between strokes. The stroke associated with the node is shown at the upper left of the node. (i.j) indicates that the stroke i of the standard pattern and the stroke j of the input pattern are associated. A five-digit number written at the bottom right of a node indicates the contents of the corresponding memory at that node.

The leftmost bit corresponds to stroke number l. For example, it indicates that strokes with stroke numbers 1, 3.4, and 10110 have already been associated.

In Figure 6, at the lower right node (33, 43), the cumulative value of the distance between strokes is 8, which is the smallest among all the nodes to be searched, and all bits in the corresponding memory are 1, so search. is completed, and the distance between the turns at this time is 8.

This inter-pattern distance is determined for all standard patterns, and is taken as the character category IJe recognition result for the standard pattern that provides the minimum inter-pattern distance. This best-priority search can be performed for all standard patterns in parallel, in addition to searching for each standard pattern as shown in this example.

In this way, in the present invention, stroke deformations such as stroke deformation due to fluctuations in the number of strokes, stroke deformation due to consecutive strokes in the wrong next stroke order, deformation of the stroke shape within one stroke, etc. The modified stroke is incorporated into the standard pattern along with the stroke number of the corresponding standard stroke. and,
By associating the strokes of the standard pattern and the strokes of the human cover pattern so that there are no surplus stroke numbers, these strokes can be matched without excess or deficiency. By recognizing characters based on the cumulative value of the distance between strokes at this time, an online character input device that does not depend on the handwriting order and has high recognition performance can be obtained. Since the correspondence of strokes does not depend on the order in which the strokes are written, this online character input device can input the character f without depending on the order of strokes. : Recognized.

(Example) An example of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.

The pattern of friend characters inputted from the tablet 1 is held in the input pattern memory 2. Standard pattern memory 3 holds standard patterns of characters to be recognized. The standard pattern includes not only standard stroke patterns but also modified stroke patterns as needed. Pattern 5 (j) of the j-th stroke of the input pattern (1≦j≦J J: number of strokes of the input pattern) is read out from the input pattern memory 2 by the control signal j,
The first stroke (1≦i≦
I tk + I fk): Standard pattern number k (
Number of strokes of DlfiA quasi-pattern. pattern M(k, i) (including modified strokes) and stroke marks l, (k, i) are read out. The stroke mark L(k,i) is data having the number of pits equal to the standard number of strokes, and the bit corresponding to the above-mentioned stroke number is 1 and the other bits are O. If there are multiple standard strokes for one modified stroke, the corresponding multiple bits are set to 1.

In the inter-stroke distance calculation unit 4, the input pattern 5(j)
Inter-stroke distance d between and standard pattern M(k,i)
(k, i, j) is calculated. This distance between strokes d (k, i, j) is based on cited document (1) I
Suppose we use Euclidean distance as mentioned in IC.

The node list memory 5 stores data Nd of the nodes of the search tree in list form in descending order of the cumulative value g(k, p) of inter-stroke distances. The cumulative value g (k, p) of the distance between strokes is. Here, 'Fri is the search depth (1≦door≦J), and T (k, j) indicates the stroke of the 2-th standard pattern that is associated with the j-th stroke of the input pattern.

The node data Nd includes the cumulative value g of the distance between strokes (
In addition to K, standard pattern number, search depth, and corresponding memo IJc(k, p) are included. In the initial state, the node list memory 5 stores data Nd of nodes for which the cumulative value g (k, p) of the distance between strokes is sufficiently large.
is retained.

Next, the node at 1)=1 is created by the node creation unit 6. and a control signal from the node creation unit 6.

When i and j are j=1.1≦ and 1≦i≦I(k), the values are sequentially output. These control signals cause the input pattern memory 2 to output the pattern SO).

From standard pattern memory 3, putter: yM (k,
i) is output, and the inter-stroke distance calculating section 4 calculates the inter-stroke distance d (k, i, j).

At the same time, the standard pattern memory 3 also outputs stroke marks L(k,i).

The node creation unit 6 creates new node data Ndnew. New node data Ndne created here
Search depth: p=1 Standard pattern number 2 k (l≦≦K) Cumulative value:
g (k, p) = d (k, i, j) Corresponding memory: C (k, I)) = L (k,
i) It is made up of the following data.

The list creation unit 7 performs a process of adding new node data Ndnew to the node list in the all node list memory 5. That is, the cumulative value g(k, p) of the distance between strokes in the data Ndnew of the new node is calculated from the beginning of the node list as the cumulative value g(k', p') of the distance between strokes of the node data Nd in the node list. As the comparison continues, new node data pJdnew is added before node data Nd that satisfies g (k, p) < g (k', p').
Add.

Subsequently, a node where p>IIC is created. The node creation unit 6 first reads data Nd2 of the node at the head of the node list in the node list memory 5 (the one with the smallest cumulative value g(k', p') of distances between strokes). Based on the data Nd of this node, a control signal if is output to the standard pattern memory 2. Here, is the standard pattern number in the node data Nd, and i is 1≦l≦I(k').
The values of i are sequentially output.

Stroke mark L (k, i
) is output, the node creation unit 6 stores the corresponding memory C(k
', p') and L (k, i) and C(k
', p') is performed. If this result is other than O, the strokes will overlap, so no further processing will be performed and the next i will be output. When the result is O, a new node Ndnew is created.

When creating a new node Ndnew, first the control signal j
is output from the node creation section 6 to the input pattern memory 2. The value of the control signal j is the search depth p' in the node data Nd.
Therefore, j=p'+1. This control signal j causes the input pattern memory 2 to output pattern SO),
To the control signals that have already been output from the six node creation units,
From iK, from standard pattern memory 3, pattern M(k
, i) are output, and the inter-stroke distance calculation unit 4 calculates the inter-stroke distance d (k, i, j).

Next, based on these values, new node data Ndn
ew is required. Namely.

Search depth: I)-1'+1 Standard pattern number: k==/ Cumulative value: g(k, p)=g(k', p')+
d(k, i, j) compatible memory: C(k-f
))-C(k', p') or L(', i). However, ortj represents a logical sum operation. This new node data Ndnew is output to the list creation section 7. The list creation unit 7 adds the new node data Ndnew to the node list in the node list memory 5 by the processing described above.

The search depth p' of the node data Nd at the head of the node list is equal to J, and the corresponding memory C(k'.

The search ends when all bits of p') are 1.

The standard pattern number k in this node data Nd is output as a recognition result.

(Effects of the Invention) According to the present invention, stroke deformations such as stroke deformation due to variations in the number of strokes, stroke deformation where strokes are continued at the wrong beginning, and stroke shape deformation within one stroke can be avoided. On the other hand, each deformed stroke is incorporated into the standard pattern, and the strokes of the standard pattern and the strokes of the input pattern can be matched with just the right amount, so the online system does not depend on stroke order and has high recognition performance even for the next deformed character. A character input device is obtained.

[Brief explanation of drawings]

Figure 1 is a configuration diagram showing one embodiment of the present invention, Figures 2 to 6
The figure is a companion diagram for explaining the present invention in detail, and FIGS. 7 and 8 are diagrams for explaining problems of the prior art. In the figure, 1...Tablet, 2...Input pattern memory, 3...Standard pattern memory, 4...
... Stroke distance calculation section, 5 ... Node list memory, 6 ... Node generation section, 7 ...
...List creation section is shown respectively. Ki 2 (2) Season 3 Illustration 4 Illustration

Claims (1)

[Claims] A handwriting detection unit that detects handwriting when writing characters and reads it as an input pattern, a standard stroke pattern of the character to be recognized, a modified stroke pattern including stroke combinations, and a stroke number and modified stroke for standard strokes. , a standard pattern memory holds the stroke numbers of one or more corresponding standard strokes together as a standard pattern, and the strokes of the input pattern and the strokes of the standard pattern are stored so that the stroke numbers do not overlap or are left over. An online character input device comprising: a recognition unit that recognizes characters based on a cumulative value of inter-stroke distances determined by associating the strokes with each other;