JPH0562392B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a handwritten character recognition device that can easily recognize handwritten characters input via a coordinate input device, regardless of the stroke order.

[Technical background of the invention and its problems]

Recently, handwritten character recognition has been developed, which uses stroke information to recognize handwritten characters entered via a coordinate input device such as a tablet device, encodes the character data indicated by the handwritten characters, and inputs the code into a computer, etc. Various devices have been proposed.

FIG. 1 shows an example.

Character patterns input by hand through a character input unit 1 consisting of a coordinate input device such as a tablet are subjected to preprocessing such as removing input noise and normalizing character size in a preprocessing unit 2. , is given to the feature extractor 3. The feature extraction section 3 extracts the features of the stroke information of the input character pattern.
The recognition processing section 5 performs a matching process between the stroke information of this input character pattern and the stroke information of each standard pattern of the recognition target category registered in advance in the standard pattern dictionary 4. This matching process is performed, for example, by calculating the degree of similarity, distance, etc. between the input pattern and each standard pattern.

The information on the similarity and distance obtained for each standard pattern in this way is evaluated by the recognition unit 6, and a recognition result (recognition target category) for the input character pattern is obtained.

By the way, the above input character pattern is, for example, the second
It is given as a series of strokes written on the tablet surface as shown in the figure. Therefore, for some scribes, the order of their written strokes (stroke order)
may differ from the normal stroke order.

In such a case, if the input stroke is matched with all the standard strokes of the standard pattern in the input order, it will be necessary to perform a very large number of matching calculations.

Therefore, conventionally, for example, as shown in Figure 3,
The standard strokes of "correct" characters and the strokes input by hand are associated on a table, and matching is performed between the strokes in the correspondence relationship. However, when the input pattern has a large number of strokes, it is very difficult to find the correspondence between strokes. Therefore, there was a problem that the recognition processing rate was low.

[Object of the Invention] The present invention has been made in consideration of the above circumstances, and its purpose is to provide a method for inputting characters by hand, regardless of the stroke input order of the character pattern.
It is an object of the present invention to provide a handwritten character recognition device that can easily and efficiently recognize the input character pattern.

[Summary of the invention]

The present invention stores characteristic information regarding the writing direction and positional coordinates of each standard stroke of a standard pattern in a standard writing stroke order in a standard pattern dictionary, and stores the writing direction and position coordinates of each stroke of an input character pattern obtained from writing input. Strokes that match the feature information regarding the position coordinates are detected in the standard writing order, and the strokes detected as matches are removed from the strokes of the input character pattern, so that all the feature information regarding the writing direction and position coordinates of the input strokes are removed. detecting a standard pattern that matches feature information regarding the writing direction and position coordinates of standard strokes; reading out the detected standard pattern from a standard pattern dictionary; and comparing it with an input character pattern in which each stroke is arranged in the standard writing stroke order; It is designed to recognize and process input character patterns.

〔Effect of the invention〕

Thus, according to the present invention, regardless of the order of input strokes, the input character pattern can be easily
Moreover, it becomes possible to recognize efficiently. Also,
Match detection between an input character pattern obtained from handwritten input and each standard stroke of a standard pattern stored in a standard pattern dictionary in the order of standard writing strokes is performed using feature information regarding the writing direction and position coordinates of the stroke. In other words, matching detection is performed not only by the position coordinates of the strokes, but also by taking into account factors such as the writing direction, so it is possible to detect an appropriate standard pattern from the matching detection of such strokes, and as a result, It is possible to realize highly accurate handwritten character recognition. In addition, while the input strokes detected as matches are removed from the strokes of the input pattern, the next matching input stroke is detected, and a standard stroke that matches all of the input strokes is detected. Even if the strokes are reordered, the strokes can be rearranged easily and reliably. Thereafter, the input character pattern is compared with the standard pattern whose strokes match, so that the amount of calculation processing can be reduced and the efficiency of the recognition processing can be improved.

[Embodiments of the invention]

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

FIG. 4 is a schematic configuration diagram of the embodiment device, and the first
The same parts as those of the conventional device shown in the figure are designated by the same reference numerals.

The feature of this embodiment device is that each standard pattern stroke feature storage unit 7, which is a part of the standard pattern dictionary, is stored in correspondence with each standard pattern of the recognition target category registered in the standard pattern dictionary 4. Register stroke order information.

The input pattern reconstruction unit 8 reconstructs the input order of the input strokes of the input pattern according to the stroke order information, and matches the input stroke order to the stroke order of the stroke standard pattern. Thereafter, a standard pattern that corresponds to all the strokes is found, and character recognition processing is performed between this standard pattern and the input pattern in which the strokes have been rearranged.

First, the basic operation of this device will be explained.

The features of the input character pattern inputted from the character input section 1 are extracted by the feature extraction section 3. The features of the input character pattern extracted by the feature extraction unit 3 include, for example, the direction of the written stroke, and the position coordinates of the starting point, middle point, and end point of each written stroke. In this case, the number of strokes of the input character pattern, that is, the number of strokes of the input character may also be extracted as feature information.

The direction of the writing stroke is defined, for example, by the slope of the straight line connecting the starting point and the ending point of each writing stroke, and specifically, as shown in FIG. It is expressed by dividing and classifying each area. Here, there are four types of writing strokes, A, B, C, and
It is classified as D. Each writing stroke is expressed using these classification types and information on the positional coordinates of the starting point, middle point, and ending point of the writing stroke.

Figures 6 and 7 show the information on the written strokes obtained for the input patterns of the characters ``ANA'' and ``明'', and each of these written strokes is numbered according to the order of writing. ing. For each writing stroke, the coordinate data of the starting point, middle point, and ending point from which the stroke was determined are written for the entire input pattern. In this example, the coordinate data is ranked in the X-axis direction and the Y-axis direction and its characteristics are expressed.

On the other hand, in the standard pattern dictionary 4, information on standard patterns is registered for each character code indicating each recognition target category, as shown in FIG. 8, for example. This standard pattern information, similar to that used in conventional recognition processing, describes the coordinate data (X, Y) of the start point, middle point, and end point of the strokes that make up each standard pattern. .

Further, the characteristic information of each standard stroke characterized by this device is stored in the standard pattern stroke characteristic storage section 7.
is registered in. The feature information of this standard stroke is described as information about the possible writing directions of the stroke. For example, when the characteristic information of the stroke is given as A and B, this indicates that the direction of the stroke is the above-mentioned writing direction A or B. Also, stroke characteristic information
When given as NA, it indicates that the direction of the stroke is other than the writing direction A.

However, the characteristic information of the standard stroke written in this manner is stored in the standard pattern stroke characteristic storage section 7 in accordance with the standard writing stroke order of the standard pattern and corresponding to that stroke.

The characteristic information of the standard strokes stored in the standard pattern stroke storage section 7 in this manner is sequentially read out in accordance with the storage order.

In the input pattern reconstruction unit 8, according to the stroke characteristic information whose order is controlled in this way,
Input strokes corresponding to each of the standard strokes in the input character pattern are sequentially detected. At this time, the detected matching strokes are removed from the strokes constituting the input character pattern, and the input strokes corresponding to the next standard stroke among the remaining strokes are sequentially detected.

Note that if an input stroke corresponding to a standard stroke is not detected, it is assumed that the standard pattern and the input pattern are different, and the process of matching strokes against the standard pattern is stopped. Stroke detection for the next standard pattern is then performed in the same manner as the stroke detection process described above.

FIG. 9 shows how the input pattern reconstruction section 8 processes the input character "correct".

Information regarding the strokes of the input characters extracted by the feature extraction section 3 is processed by the input pattern reconstruction section 8 and sent to the recognition processing section 5.

That is, the information regarding the stroke of the input pattern extracted by the feature extraction section 3 is first supplied to the input buffer of the reconstruction section 8 and stored. The reconstruction unit 8 checks whether the input character pattern has a stroke corresponding to the stroke characteristic of the character "correct" shown in FIG. 9, for example.

That is, the matching section first searches for a stroke that corresponds to the first standard stroke feature from among the input strokes stored in the input buffer. in this case,
Among the "positive" input strokes, there is a stroke indicated by the dashed line in FIG. On the other hand, the verification section deletes the stroke data from the input buffer.

Next, the matching unit searches the input buffer again for a stroke that corresponds to the second feature. In this case, the stroke shown by the dashed line in FIG. 9b is detected from among the strokes remaining in the input buffer, so this stroke data is written to the output buffer as the second order stroke data, and , delete the relevant stroke data from the input buffer.

Further, the matching unit continues searching in the same way. That is, strokes corresponding to the third characteristic are checked. At this time, among the remaining strokes, the stroke shown by the dashed line in FIG. 9c is detected. Therefore, the collation section writes the data of this stroke as the stroke data of the third order in the output buffer, and deletes the data of the stroke from the input buffer.

In this way, in this example, since it is known that the input character pattern consists of five strokes (five strokes), strokes corresponding to the characteristics of the first to fifth standard strokes are sequentially detected. As a result, the output buffer is formed with input stroke data rearranged according to the standard stroke order.

The above is the basic operation of the input pattern reconstruction section 8.

In this way, the recognition processing unit 5 compares the input strokes associated with the stroke order of the standard pattern with the strokes of the standard pattern.

This comparison compares the shape of the stroke indicated by the coordinate positions of the start point, midpoint, and end point and the positional relationship of the stroke with respect to the character between the input stroke and the standard stroke, and calculates the degree of similarity and distance between them. This is done by calculating . The input pattern is recognized by evaluating the similarity of the input pattern to the standard pattern based on the degree of similarity and the distance.

Note that the process of comparing the standard stroke and the input stroke may be performed after limiting the number of strokes of the standard pattern using, for example, information on the number of strokes of the input pattern. In this way, there is no need to perform matching processing on standard patterns with different numbers of strokes, so it is possible to eliminate waste in the processing process and improve recognition processing efficiency.

As described above, according to the present device, the order of input strokes is rearranged, and only standard patterns having standard strokes that match the input strokes are selected and compared with the input pattern, so that the input pattern can be written easily. Regardless of the stroke order and regardless of the number of strokes,
Input character patterns can be recognized efficiently.

In addition, a match between the input character pattern obtained from the handwritten input and each standard stroke of the standard pattern stored in the standard pattern dictionary in the order of standard writing strokes is detected using feature information regarding the writing direction and position coordinates of the stroke. In other words, matching detection is performed not only by the position coordinates of the strokes, but also by taking into account factors such as the writing direction, so it is possible to detect an appropriate standard pattern from the matching detection of such strokes, and this As a result, highly accurate handwritten character recognition can be achieved.

Therefore, it is possible to simplify the input format without placing an extra burden on the scribe regarding the order of input strokes. In other words, even if you input the input character pattern in the wrong stroke order,
It becomes possible to recognize the input character pattern, thereby reducing input errors and improving input efficiency.

Furthermore, if there is no input stroke that corresponds to a standard stroke, the recognition process for that standard stroke is stopped at that point and the process moves on to the next standard pattern, so there is no waste in the recognition process. Therefore, input patterns can be recognized efficiently in a short time, and tremendous practical effects such as improved processing efficiency can be achieved.

Note that the present invention is not limited to the embodiments described above. For example, feature information of a standard stroke of a standard pattern may be provided as a more detailed feature. That is, the direction of the stroke may be classified more finely, and the position information thereof may also be added to detect the stroke. Further, the characteristics of the stroke may be provided as information on the position from the coordinate origin and the direction of writing of the stroke. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of a conventional character recognition device, FIG. 2 is a diagram showing an example of an input pattern input by handwriting, FIG. 3 is a diagram showing an example of conventional stroke matching, and FIG. 4 is a diagram showing the present invention. 5 is a diagram showing an example of classification of stroke directions showing characteristics of strokes; FIGS. 6 and 7 are diagrams showing input patterns and characteristic information of the input strokes, respectively; FIG. 8 is a diagram showing the structure of the standard pattern dictionary, and FIG. 9 is a diagram showing the flow of stroke detection by this device. 1...Character input unit, 2...Preprocessing unit, 3...Feature extraction unit, 4...Standard pattern dictionary, 5...Recognition processing unit, 6
...Identification section, 7. Standard pattern stroke feature storage section, 8. Input pattern reconstruction section.

Claims (1)

[Scope of Claims] 1. A standard pattern dictionary that stores feature information regarding the writing direction and position coordinates of each standard stroke of this standard pattern in the order of standard writing strokes, together with standard patterns of character categories to be recognized; and characters that have been input by hand. means for detecting feature information regarding the writing direction and positional coordinates of each stroke of the pattern; and means for sequentially reading out the feature information regarding the writing direction and positional coordinates of the standard strokes stored in the standard pattern dictionary, and detecting feature information regarding the writing direction and positional coordinates of each stroke of the pattern; Detecting strokes whose feature information regarding the writing direction and position coordinates match the standard stroke in each stroke in the standard writing order,
The detected strokes are removed from the strokes of the input character pattern to detect a standard pattern in which all of the characteristic information regarding the writing direction and positional coordinates of the input stroke match the characteristic information regarding the writing direction and positional coordinates of the standard stroke. and means for reading the detected standard pattern from the standard pattern dictionary, comparing it with the input character pattern in which each stroke is arranged in the standard writing stroke order, and recognizing and processing the input character pattern. A handwritten character recognition device characterized by: 2. If the standard pattern detection means cannot find a stroke of the input pattern that corresponds to one standard stroke in the standard pattern, the standard pattern detection means stops processing the standard pattern. Claim 1 The handwritten character recognition device described. 3. The handwritten character recognition device according to claim 1, wherein the recognition process by matching the standard pattern and the input pattern is obtained by matching the standard pattern and the input pattern.