Handwritten character verification method and apparatus therefor

• US 6,160,914 A
• Filed: 12/11/1997
• Issued: 12/12/2000
• Est. Priority Date: 11/08/1996
• Status: Expired due to Fees

• Current Assignee: CADIX, Inc.
• Original Assignee: CADIX, Inc.
• Inventors: Muroya, Masami
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): DESIRE, GREGORY M
• Application Number:
  US08/989,081
• Time in Patent Office: 1,097 Days
• Field of Search: 382/119, 382/120, 382/121, 382/122, 382/123, 382/187, 382/18, 382/209, 382/215, 382/218
• US Class Current: 382/187
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures

Abstract

A handwritten character recognition method including the steps of acquiring coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, correcting the handwritten character string data through use of a warping function in order to minimize the overall degree of difference among a plurality of sets of handwritten character string data, and comparing the thus-corrected character string data with registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the handwritten character string data and the corrected handwritten character string data. The method further includes the steps of extracting strokes from the character string, acquiring coordinate information and writing pressure information with regard to each of the strokes corresponding to the character string to be checked, and summing cumulative errors between corresponding paired strokes of the character strings to be compared in terms of coordinate information and writing pressure information, to thereby calculate from the cumulative values the overall degree of difference between the handwritten character string data and registered handwritten character string data. Further, the degree of difference in fluctuation of writing movement is calculated from cumulative values of the fluctuation speed of writing movement through use of time axis warping functions. These degrees of difference are added to discriminant elements.

Claims

1. A handwritten character recognition method by which the matching of characters is determined by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the method comprising the steps of:

acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information;

correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data;

comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

extracting characters from the registered handwritten character string data and the input handwritten character string data;

normalizing the thus-extracted characters on a character-by-character basis with regard to position and size;

calculating cumulative errors in coordinate and a cumulative error in writing pressure between the registered handwritten character string data and the input handwritten character string data for each character;

summing the cumulative errors for each character to thereby produce the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data; and

determining whether the handwritten character string is genuine or false from the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, as well as from the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data.

2. A handwritten character recognition method by which the matching of characters is determined by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the method comprising the steps of:

acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information;

correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data;

comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

calculating the strength of fluctuation of writing movement from the registered handwritten character string data and the input handwritten character string data as a rate of change in fluctuation through use of time axis warping functions;

calculating a cumulative value of the rate of change in fluctuation to thereby obtain the degree of difference in fluctuation of writing movement; and

determining whether the handwritten character string is genuine or false from the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, as well as from the degree of difference in fluctuation of writing movement.

3. A handwritten character recognition method by which the matching of characters is determined by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the method comprising the steps of:

acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information;

correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data;

comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

extracting characters from the registered handwritten character string data and the input handwritten character string data;

normalizing the thus-extracted characters on a character-by-character basis with regard to position and size;

calculating a cumulative error in coordinate and a cumulative error in writing pressure between the registered handwritten character string data and the input handwritten character string data on a character-by-character basis;

summing the cumulative errors of the respective characters to thereby produce the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data;

calculating the strength of fluctuation of writing movement from the registered handwritten character string data and the input handwritten character string data as a rate of change in fluctuation through use of time axis warping functions;

calculating a cumulative value of the rate of change in fluctuation to thereby obtain the degree of difference in fluctuation of writing movement; and

determining whether the handwritten character string is genuine or false from the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, from the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data, as well as from the degree of difference in fluctuation of writing movement.

4. A handwritten character recognition apparatus in which the matching of characters is determined by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the apparatus comprising:

means for acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information, correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data, and comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

means for extracting characters from the registered handwritten character string data and the input handwritten character string data, normalizing the thus-extracted characters on a character-by-character basis with regard to position and size, calculating cumulative errors in coordinate and a cumulative error in writing pressure between the registered handwritten character string data and the input handwritten character string data for each character, and summing the cumulative errors for each character to thereby produce the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data, whereby a determination is made as to whether the handwritten character string is genuine or false, on the basis of the degree of difference between the corrected handwritten character string data and the registered handwritten character string data, as well as of the overall degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data.

5. A handwritten character recognition apparatus in which the matching of characters is determined by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the apparatus comprising:

means for acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information, correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data, and comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data; and

means for calculating the strength of fluctuation of writing movement from the registered handwritten character string data and the input handwritten character string data as a rate of change in fluctuation through use of time axis warping functions and calculating a cumulative value of the rate of change in fluctuation to thereby obtain the degree of difference in fluctuation of writing movement, whereby a determination is made as to whether the handwritten character string is genuine or false, on the basis of the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, as well as of the degree of difference in fluctuation of writing movement.

6. A handwritten character recognition apparatus in which the matching of characters is determined by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the apparatus comprising:

means for acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information, correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data, and comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

means for extracting characters from the registered handwritten character string data and the input handwritten character string data, normalizing the thus-extracted characters on a character-by-character basis with regard to position and size, calculating cumulative errors in coordinate and a cumulative error in writing pressure between the registered handwritten character string data and the input handwritten character string data for each character, summing the cumulative errors for each character to thereby produce the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data; and

means for calculating the strength of fluctuation of writing movement from the registered handwritten character string data and the input handwritten character string data as a rate of change in fluctuation through use of time axis warping functions and calculating a cumulative value of the rate of change in fluctuation to thereby obtain the degree of difference in fluctuation of writing movement, whereby a determination is made as to whether the handwritten character string is genuine or false, on the basis of the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, of the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data, as well as of the degree of difference in fluctuation of writing movement.

7. A medium on which recorded is a program for determining the matching of characters by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the medium comprising the procedures for:

acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information;

correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data;

comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

extracting characters from the registered handwritten character string data and the input handwritten character string data, normalizing the thus-extracted characters on a character-by-character basis with regard to position and size;

calculating cumulative errors in coordinate and a cumulative error in writing pressure between the registered handwritten character string data and the input handwritten character string data for each character;

summing the cumulative errors for each character to thereby produce the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data; and

making a determination as to whether the handwritten character string is genuine or false, on the basis of the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, as well as of the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data.

8. A medium on which recorded is a program for determining the matching of characters by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the medium comprising the procedures of:

acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information;

correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data;

comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

calculating the strength of fluctuation of writing movement from the registered handwritten character string data and the input handwritten character string data as a rate of change in fluctuation through use of time axis warping functions;

calculating a cumulative value of the rate of change in fluctuation to thereby obtain the degree of difference in fluctuation of writing movement; and

determining whether the handwritten character string is genuine or false from the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, as well as of the degree of difference in fluctuation of writing movement.

9. A medium on which recorded is a program for determining the matching of characters by acquisition of coordinate information and writing pressure information relating to a handwritten character string as three-dimensional time series information, and by comparison of the acquired three-dimensional time series information with handwritten character string data registered beforehand, the method comprising the steps of:

acquiring coordinate information and writing pressure information relating to the handwritten character string as three-dimensional time series information;

correcting the handwritten character string data through use of a warping function in order to minimize a cumulative error between the handwritten character string data and the registered handwritten character string data;

comparing the thus-corrected character string data with the registered handwritten character string data by dynamic programming matching so as to obtain the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data;

extracting characters from the registered handwritten character string data and the input handwritten character string data;

normalizing the thus-extracted characters on a character-by-character basis with regard to position and size;

calculating cumulative errors in coordinate and a cumulative error in writing pressure between the registered handwritten character string data and the input handwritten character string data for each character;

summing the cumulative errors for each character to thereby produce the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data;

calculating the strength of fluctuation of writing movement from the registered handwritten character string data and the input handwritten character string data as a rate of change in fluctuation through use of time axis warping functions;

calculating a cumulative value of the rate of change in fluctuation to thereby obtain the degree of difference in fluctuation of writing movement; and

determining whether the handwritten character string is genuine or false from the overall degree of difference between the corrected handwritten character string data and the registered handwritten character string data, from the degree of character-by-character difference between the corrected handwritten character string data and the registered handwritten character string data, as well as from the degree of difference in fluctuation of writing movement.