JPS62286190A - Identification system for online voiced sound and semi-voiced sound - Google Patents

Abstract

PURPOSE:To correctly identify characters having a voiced sound and a semi- voiced sound by using comparatively stable characteristic of the position of the coordinates of feature points such as the start point, the end point and the middle point of a stroke in an online hand writing character recognition. CONSTITUTION:A recognition candidate is decided by a general sorting circuit 2 for a character pattern hand writing inputted to a tablet 1. In the case of the character having the voiced sound and the semi-voiced sound, the coordinate of the feature point detected by a feature point detecting circuit 4 and the description in a feature point dictionary 3 are compared in an identification circuit 5. As the result of this, the voiced sound or the semi-voiced sound is decided and a final recognition result is decided. For instance, if the recognition candidate of the input pattern is 'pa' or 'ba' by the circuit 2, a detailed identification is carried out according to the contents of the dictionary 3. The position of a stroke second from the last is detected by a circuit 4 and this position is not a prescribed right upper position, so that it is decided that the input pattern is 'pa' by a circuit 5.

Description

[Detailed Description of the Invention] 3. Detailed Description of the Invention [1 Already Required] The present invention solves the problem in online handwritten character recognition, where it becomes difficult to distinguish between voiced and semi-voiced sounds due to changes in the number of strokes of characters and changes in the shape of loops. In order to solve this problem, by using relatively stable features such as the coordinate positions of feature points such as the start point, end point, and midpoint of a stroke, characters with voiced sounds and half-voiced sounds can be correctly identified. This is what I did.

[Fields of use for graduation]

The present invention relates to online handwritten character recognition in which characters are handwritten into a tablet and recognized in real time, and particularly relates to the identification of characters with voiced and half-voiced sounds in hiragana and katakana.

[Conventional technology]

Conventional methods for identifying voiced sounds and semi-voiced sounds include: (1) identification based on the number of strokes of an input character, and (2) identification by detecting loops of semi-voiced sounds. FIG. 4 is a block diagram showing an outline of the former (2). For characters input by hand on the tablet 6, recognition candidates are determined by the major classification circuit 7, but if a character with a voiced or semi-voiced sound becomes a recognition candidate, detailed recognition is performed by the stroke count determination circuit 8. . If the character recognized as "Ba" by the major classification method has 4 strokes, it becomes "Ba" and 5.
If it is a picture, it will be recognized as "ba". FIG. 5 shows an outline of the latter method (2), and the difference from FIG. 4 is that a loop detection circuit 1) is used instead of the stroke number determination circuit 8. A character recognized as "Ba" by the major classification method is recognized as "Ba" if a loop is detected from the final stroke, and as "Ba" if no loop is detected.

[Problem that the invention seeks to solve]

In the conventional voiced and semi-voiced sound identification method, which identifies characters based on the number of strokes of an input character, there is a problem in that a pattern with a variation in the number of strokes, such as "wa 0", is misread.

In addition, in a method that performs identification by detecting loops, "1
When the shape of the loop changes drastically like "v", there is a problem that the loop is not ejected.

[Means for solving problems]

FIG. 1 is a principle block diagram of an online handwritten character recognition device using the voiced and semi-voiced sound discrimination method of the present invention. In the figure, 1 is a character input tablet, 2 is a major classification circuit, 3 is a feature point dictionary that describes the positions of feature points of strokes of characters with voiced or semi-voiced sounds, and 4 is a feature point of strokes. A feature point detection circuit 5 detects the position of the coordinates, and a discrimination circuit 5 identifies voiced sounds and semi-voiced sounds from the results of the feature point detection circuit 4 and the descriptions in the feature point dictionary 3.

[Effect]

The character pattern handwritten into the tablet 1 shown in FIG. The coordinates of the detected feature point and the description in the feature point dictionary 3 are compared in the identification circuit 5,
The final recognition result is determined by determining whether the sound is voiced or semi-voiced.

〔Example〕

The operation of the embodiment will be explained with reference to the principle diagram shown in FIG.

FIG. 2 is an example of a handwritten cover pattern, and FIG. 3 is a diagram showing part of the contents of the feature point dictionary 3.

For example, assume that the pattern shown in FIG. 2 is input by hand on the tablet 1. Conventional methods that use stroke count information cannot correctly recognize this character because it only has three strokes.
Furthermore, in the second conventional method that uses loop detection, loops are not detected due to variations in the shape of the loops. on the other hand,
In the present invention, if the major classification circuit 2 in the configuration of FIG. 1 determines that the recognition candidate for the input pattern in FIG. Detailed identification is performed according to the contents of 3. The feature point detection circuit 4 detects the position of the end point of the second to last stroke, and since this position is not at the upper right, the identification circuit 5 determines that the incoming cover turn is 7B.

The above example shows the case of the character pair "ba, ba", but other character pairs "bi, bi", "bu, bu", "be, pe", "bo, bo" It goes without saying that the present invention can be similarly applied to other cases.

[Effects of the Invention] According to the present invention, it is possible to stably identify voiced sounds and semi-voiced sounds even in characters with variations in the number of strokes and variations in the shape of loops, so there are fewer restrictions on scribes and an easier-to-use system can be provided. be.

[Brief explanation of drawings]

Fig. 1 is a diagram of the principle of the present invention, Fig. 2 is an example of a handwritten cover pattern, Fig. 3 is a diagram showing part of the contents of a feature point dictionary, and Fig. 4 is the principle of the conventional method of identification based on the number of strokes. FIG. 5 is a diagram showing the principle of a conventional method that performs identification by detecting loops. In Figure 1, 1 is a tablet, 2 is a major classification circuit, and 3 is a tablet.
4 is a feature point dictionary, 4 is a feature point detection circuit, and 5 is an identification circuit. Now to bitterness Kabataan f4Ii Rime 2 Figure @ Similarity 1/) Jikure 1 also 3 ■Ha

Claims (2)

[Claims] (1) In an online handwritten character recognition device, a feature point dictionary (3) that describes the positions of the feature points of the strokes of kana characters with voiced and semi-voiced sounds and feature point detection that detects the coordinate positions of the stroke feature points. Means (4); Means for identifying voiced sounds and semi-voiced sounds of kana characters based on the coordinate positions of the feature points detected by the feature point detection means (4) and the contents of the feature point dictionary (3); 5) An online voiced sound and semi-darkened sound identification method characterized by having the following. (2) A patent characterized in that when the end point position of the penultimate stroke is located at the upper right corner of the character, it is determined to be a voiced sound, and when the end point position is located at a position other than the upper right corner of the character, it is determined to be a semi-voiced sound. An online voiced sound and semi-darkened sound identification method according to claim (1).