JPS6022793B2 - character identification device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character identification device for identifying handwritten characters online.

Character identification devices include so-called OCR, which performs identification by photoelectrically converting characters written in advance on a form, etc., and online identification, which performs identification by inputting the process of pen movement while writing characters.

The former has already been put to practical use and is widely used, but the latter has not yet been put into practical use, despite its advantage of being able to be encoded immediately at the data source. The reason for this is that in the latter case, the classification efficiency for selecting candidate characters for identification based on the presence or absence of strokes, stroke order, etc. is low, and as the number of characters increases, the manpower and time required for dictionary design increases. It is thought that there is. The object of the present invention is to provide a character identification device which eliminates the above-mentioned drawbacks.

FIG. 1 shows a block diagram of an embodiment of the present invention.

1 is a data tablet, 2 is a pen for character input, 3 is a character input section, 4 is a character synthesis section, 5 is a preprocessing section, 6 is a collation section,
7 is a dictionary, and 8 is a determination section.

Among these, the data tablet 1, Ben 2, and character input section 3 have the same configuration as the input section of a normal online identification device, and by drawing the character you want to input on the Ben 2 data tablet 1, you can In addition to data on the drawn characters, information such as stroke order, number of strokes, direction and speed of Ben 2's movement (hereinafter referred to as stroke information) is obtained. Here, as shown in FIG. 2, the input data includes the position coordinates (X, Y) of the pen 2 at time m sampled at regular intervals and the difference ΔX, Δ between the coordinates at time tn-. Y is obtained. This data sampling is performed while the pen is in contact with the data tablet 1, and it is assumed that there is no change in the data when the pen 2 is not in contact with the data tablet.
The above-mentioned stroke information can also be obtained from these data. The character synthesis section 4 synthesizes a character pattern from the data obtained by the character input section 3.

That is, the character synthesis section 4 has a pattern memory (not shown), and this pattern memory is set to 0 in advance. Next, each false coordinate (x, Y) sampled in the input order shown in FIG. 2 is read out, and the corresponding point (X, Y) on the pattern memory is set to 1. In this way, by writing 1 on the pattern memory corresponding to all the input coordinates, an input character pattern sampled in pieces is formed on the pattern memory. By subjecting this character pattern to connection processing using the above-mentioned stroke information, an input character pattern in which each sample point is connected is obtained. This input character pattern is sent to the preprocessing section 5, subjected to known size normalization and line thickening processing, and then supplied to the collation section 6. The matching unit 6 calculates the similarity of character input patterns using a well-known pattern matching method.

That is, the dictionary 7 stores a large number of standard patterns, and the degree of similarity between each of them and the character input pattern is calculated, and the standard pattern whose degree of similarity is equal to or greater than a predetermined value is selected as the character input pattern. is determined as a candidate character, and is notified to the determination unit 8. The determination unit 8 determines the character to which the input character belongs from among these candidate characters based on the stroke information. For example, assume that the matching unit 6 determines "so" and "n" as candidate characters as shown in FIGS. 3a and 3b.

Although it is difficult to distinguish between the two using pattern matching methods,
According to the stroke information, the first stroke 10 is almost the same, but the second stroke 11 is in the opposite direction, so it can be easily distinguished. As described above, according to the present invention, handwritten characters ranging from Chinese characters to alphanumeric characters can be identified online with high accuracy.

In this case, the time required for recognition depends on the number of categories to be identified, but in OCR, devices with performance of 100 U/sec or more can be realized relatively easily, and in the case of online recognition, character resynthesis is required. There is no problem in practical terms if the sand is about 1/1 of the difference between the process and the identification. Therefore, it is also possible to realize a system having a plurality of character input devices such as data tablets. In addition, in the above embodiment, candidate characters were determined by the matching section using the pattern matching method.
Conversely, a configuration may be adopted in which candidate characters are determined from stroke information and the determination section performs determination using a pattern matching method.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the invention, and FIGS. 2 and 3 a and 3b are diagrams for explaining the operation of the embodiment of the invention. 1... Data tablet, 2... Pen, 3... Character input unit, 4... Character synthesis unit, 5... Preprocessing unit, 6... Collation unit, 7...Dictionary, 8.
... Judgment department. Figure 1 Figure 2 Figure 3

Claims (1)

[Claims] 1. Character input means that observes the movement of the pen point of an input character, obtains the position of the pen point as a time series of coordinate values, and extracts its stroke information, and uses the coordinate value string obtained from this character input means. a character processing means for forming a geometric character pattern; a storage means for storing a large number of geometric standard character patterns in advance; a collation means that calculates the degree of similarity by comparing the character pattern with the character pattern and outputs the standard character pattern as an input character candidate based on this degree of similarity; and if there are multiple standard character patterns output from the collation means, the character input means; 1. A character identification device comprising: determination means for determining a standard character pattern corresponding to an input character using stroke information extracted by the method.