Image analysis – Pattern recognition – Unconstrained handwriting
Reexamination Certificate
2000-08-23
2004-04-20
Chang, Jon (Department: 2623)
Image analysis
Pattern recognition
Unconstrained handwriting
C382S187000, C382S217000
Reexamination Certificate
active
06724936
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to input devices using handwriting recognition.
BACKGROUND OF THE INVENTION
Various methods for handwriting recognition (HWR) have been disclosed in the literature. Some of them categorize characters based on general databases containing all possible characters. The drawback of this approach is that the corresponding algorithms are very complicated and comparatively slow. This is because it is difficult to achieve good recognition accuracy using simple algorithms. Other methods, such as those disclosed in U.S. Pat. Nos. 4,531,231, 5,889,888, 5,903,667, 6,055,333, use a plurality of different input areas for inputting different character sets such as alphabetic and numeric characters. The different character sets are input into their respective input areas in order to simplify recognition algorithms and to increase recognition probability. The idea is that the algorithms for each input area can be limited to features of the corresponding character set. This approach, however, is not always feasible. There are devices, cellular phones, for example, where input areas are purposely made as small as possible, and it is impossible or inconvenient, to use a plurality of input areas.
SUMMARY OF THE INVENTION
There is provided, in accordance with a preferred embodiment of the present invention, a method for recognizing input characters handwritten onto an input area of an input device, each point of the input area being represented by coordinate values. The method includes the steps of receiving a signal representative of an input character, determining an input position representative of the input character using the information about the coordinate values, and recognizing the input character as one of a plurality of reference characters in a reference library. The recognizing step at least uses recognition features and a weighting function of each of the plurality of reference characters in the reference library and the input position. The signal includes at least information about the coordinate values of the input character.
Furthermore, in accordance with a preferred embodiment of the present invention, the input position is a center of gravity of the input character.
Additionally, in accordance with a preferred embodiment of the present invention, the coordinate values include a first coordinate value and a second coordinate value and the input position is at least one of a maximal value of the first coordinate value, a maximal value of the second coordinate value, a minimal value of the first coordinate value, and a minimal value of the second coordinate value.
Moreover, in accordance with a preferred embodiment of the present invention, the coordinate values include a first coordinate value and a second coordinate value and the weighting function is a linear function or a non-linear function of at least one of the first coordinate value and the second coordinate value.
Additionally, in accordance with a preferred embodiment of the present invention, the weighting function is a constant determined from a look-up table.
Moreover, in accordance with a preferred embodiment of the present invention, the plurality of reference characters represents a single character set.
Furthermore, in accordance with a preferred embodiment of the present invention, the reference library is a single library,
Additionally, in accordance with a preferred embodiment of the present invention, the step of recognizing further includes the steps of extracting input recognition features of the input character, comparing between the input recognition features and the recognition features of the reference characters and generating a primary recognition probability vector, determining a position probability vector for the input character using the input position and the weighting function of each of the plurality of reference characters, determining a general recognition probability vector for the input character using the primary recognition probability vector and the position probability vector, and selecting a reference character having the highest general recognition probability vector as the recognized character.
There is also provided, in accordance with a preferred embodiment of the present invention, a method for distinguishing among handwritten characters. The method includes the steps of receiving information about an input character including at least one input position, having a library of reference characters, and recognizing the input character as one of the reference characters by at least using the input position and the reference position information. Each one of the reference characters has reference position information associated with it.
Moreover, in accordance with a preferred embodiment of the present invention, the at least one input position is a center of gravity of the input character.
Furthermore, in accordance with a preferred embodiment of the present invention, the at least one input position includes a first coordinate value and a second coordinate value.
Additionally, in accordance with a preferred embodiment of the present invention, the at least one input position is at least one of a maximal value of the first coordinate value, a maximal value of the second coordinate value, a minimal value of the first coordinate value, and a minimal value of the second coordinate value.
Moreover, in accordance with a preferred embodiment of the present invention, the reference position information is a linear or a non-linear function of at least one of the first coordinate value and the second coordinate value.
Additionally, in accordance with a preferred embodiment of the present invention, the reference position information is a constant determined from a look-up table.
Furthermore, in accordance with a preferred embodiment of the present invention, the library represents a single character set.
Moreover, in accordance with a preferred embodiment of the present invention, the library is a single library.
Still further, in accordance with a preferred embodiment of the present invention, the reference position information includes a probability function of position.
Moreover, in accordance with a preferred embodiment of the present invention, the step of recognizing includes the steps of extracting input features from the input character, generating a primary recognition probability using the input features and reference features of the reference characters, generating a position probability using the input position and the reference position information, generating a general probability for each of the reference characters using the primary recognition probability and the position probability, and selecting the reference character whose general probability is best according to a selection criteria.
There is provided, in accordance with a preferred embodiment of the present invention, a system for the recognition of input characters handwritten onto an input area of an input device, each point of the input area being represented by coordinate values. The system includes a unit for receiving a signal representative of an input character, a unit for determining an input position representative of the input character using the information about the coordinate values, and a handwriting recognition unit for recognizing the input character as one of a plurality of reference characters in a reference library. The handwriting recognition unit at least uses recognition features and a weighting function of each of the plurality of reference characters in the reference library and the input position. The signal includes at least information about the coordinate values of the input character.
Additionally, in accordance with a preferred embodiment of the present invention, the handwriting recognition unit includes a module for extracting input recognition features of the input character, a module for comparing between the input recognition features and the recognition features of the reference characters and generating a primary recognition probab
ART-Advanced Recognition Technologies, Ltd.
Chang Jon
Eitan, Pearl, Latzer & Cohen Zedek LLP
LandOfFree
Handwriting input device and method using a single character... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Handwriting input device and method using a single character..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Handwriting input device and method using a single character... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3257098