Image analysis – Pattern recognition – On-line recognition of handwritten characters
Reexamination Certificate
1999-04-26
2001-09-11
Bella, Matthew C. (Department: 2621)
Image analysis
Pattern recognition
On-line recognition of handwritten characters
C382S190000, C382S197000, C382S209000
Reexamination Certificate
active
06289124
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to methods and systems of recognizing handwritten characters.
2. Description of the Background Art
In recent years a pen rather than a keyboard has been used to input characters in portable information equipment and the like. For example, for a tablet with display and input devices integrated therein that is mounted, e.g., in a PDC, recognition rate of handwritten characters input thereto is an important factor in evaluating the value thereof as a product. Methods of such character recognition include comparing stroke coordinates of a handwritten character with a dictionary database, comparing the writing direction of a handwritten character as well as its originating order with a dictionary database.
In the former method of character recognition, all stroke coordinates must be subjected to the comparison. Thus an enormous amount of information must be processed particularly for characters which vary in stroke order and stroke number. This disadvantageously requires a large capacity of memory for the dictionary database and processing programs and also results in low processing speed.
In the latter method of character recognition, more specifically, the direction of each stroke of a handwritten character is detected with respect to the time at which the user starts to write it (or its originating order) and the detected direction is compared with the dictionary database in memory. The character in the dictionary database that best matches the input handwritten character is determined as a recognized character. While the latter method requires smaller memory capacity for the dictionary database and also provides faster processing speed than the former method, the character recognition rate of the latter method is significantly reduced for a character written in an incorrect stroke order, since the latter method refers to the writing time of the character to detect the writing direction of the character.
Japanese Patent Laying-Open No. 10-154206 discloses a method in which the level of a feature of each stroke's turn in a handwritten character is detected and compared with a dictionary database to identify the character. In accordance with this method the position and angle of each stroke's turn of a character are detected and compared with the dictionary database to identify the character. Thus the amount of information required for character recognition can be reduced and characters written in incorrect stroke orders can also be recognized correctly. However, the character recognition rate of this method is also degraded for cursive characters, since cursive characters have irregular turns caused depending on how the characters are written so that the positions of the turns cannot be detected correctly. This disadvantage is similarly found in the above-mentioned methods of comparing stroke coordinates and comparing writing directions.
SUMMARY OF THE INVENTION
One object of the present invention is to provide a character recognition method and system capable of rapidly and highly reliably recognizing characters including cursive characters while reducing the amount of information required for character recognition.
In one aspect of the present invention a method of recognizing input handwritten characters includes the steps of: linking from the ending point of each actual stroke of an input handwritten character to the starting point of the subsequent actual stroke of the input handwritten character with an imaginary stroke to form a single line; detecting the level of a feature for specifying a turn of the single line; comparing the detected feature level to each feature level of a plurality of handwritten characters in a previously prepared dictionary database to calculate the resemblance of the input handwritten character to each handwritten character in the dictionary database; and determining as the input handwritten character the handwritten character in the dictionary database having the closest, calculated resemblance. The feature preferably includes the position of the turn and still preferably includes the direction and angle of the turn.
In another aspect of the present invention a system recognizing input handwritten characters includes: a device forming a single line, linking from the ending point of each actual stroke of an input handwritten character to the starting point of the subsequent actual stroke of the input handwritten character with an imaginary stroke; a device detecting the level of a feature for specifying a turn of the single line formed by the forming device; a dictionary memory storing the levels of the feature of a plurality of handwritten characters of a dictionary database formed previously; a device comparing the feature level detected by the detecting device to each feature level of the plurality of handwritten characters in the dictionary database stored in the dictionary memory to calculate the resemblance of the input handwritten character to each handwritten character in the dictionary database; and a device determining as the input handwritten character the handwritten character in the dictionary database having the closest resemblance calculated by the comparing device. The feature preferably includes the location of the turn and still preferably includes the direction and angle of the turn.
The system still preferably includes a temporary memory including a plurality of spatial, input-feature storage regions. The plurality of spatial, input-feature storage regions are associated with a plurality of predetermined reference directions and store a feature level represented in directional change for specifying the angle of a turn. The dictionary memory includes a plurality of spatial, dictionary data feature storage regions associated with the plurality of reference directions and storing a feature level in directional change for specifying the angle of a turn of a handwritten character in the dictionary database. The detecting device stores a feature level in directional change into the temporary memory at the address corresponding to the detected position of a turn. The comparing device compares each spatial, input-feature storage region of the temporary memory with the associated spatial, dictionary data feature storage region of the dictionary memory.
Preferably the detecting device assumes both two of the plurality of reference directions adjacent to the direction of a turn as the direction of the turn. Still preferably the detecting device weights feature levels in directional change depending on the proximity of the both adjacent reference directions to the direction of the turn and stores the weighted feature levels in directional change into spatial, input-feature storage regions respectively associated with the both adjacent reference directions.
The above method and system link actual strokes of a handwritten character together using imaginary strokes and then detect the positions of turns in the character which are in turn compared with a dictionary database to identify the character. Thus the position of a turn of a cursively handwritten input character can also readily be associated with that of a turn of a handwritten character in the dictionary database to increase recognition rate. In addition to turn position, turn direction and turn angle can also be detected to further increase recognition rate.
Furthermore, turn direction can be normalized with respect to a plurality of predetermined reference directions to reduce the amount of information to be stored in the dictionary memory and also increase processing speed.
The foregoing and other objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings.
REFERENCES:
patent: 4653107 (1987-03-01), Shojima et al.
patent: 4757551 (1988-07-01), Kobayashi et al.
patent: 4813078 (1989-03-01), Fujiwara et al.
patent: 4989258 (1991-01-01),
Armstrong Westerman Hattori McLeland & Naughton LLP
Bella Matthew C.
Desire Gregory
Sanyo Electric Co,. Ltd.
LandOfFree
Method and system of handwritten-character recognition does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system of handwritten-character recognition, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system of handwritten-character recognition will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2545620