Key character extraction and lexicon reduction for cursive text recognition

• US 6,249,605 B1
• Filed: 09/14/1998
• Issued: 06/19/2001
• Est. Priority Date: 09/14/1998
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Zimmerman, Matthias, Mao, Jianchang
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): Dastouri, Mehrdad
• Application Number:
  US09/152,587
• Time in Patent Office: 1,009 Days
• Field of Search: 382/155, 382/156, 382/159, 382/160, 382/161, 382/177, 382/178, 382/179, 382/186, 382/187, 382/188, 382/189, 382/119, 382/123, 706/25
• US Class Current: 382/186
• CPC Class Codes: G06V 30/2272 with lexical matching

Abstract

A method, apparatus, and article of manufacturing employing lexicon reduction using key characters and a neural network, for recognizing a line of cursive text. Unambiguous parts of a cursive image, referred to as “key characters,” are identified. If the level of confidence that a segment of a line of cursive text is a particular character is higher than a threshold, and is also sufficiently higher than the level of confidence of neighboring segments, then the character is designated as a key character candidate. Key character candidates are then screened using geometric information. The key character candidates that pass the screening are designated key characters. Two-stages of lexicon reduction are employed. The first stage of lexicon reduction uses a neural network to estimate a lower bound and an upper bound of the number of characters in a line of cursive text. Lexicon entries having a total number of characters outside of the bounds are eliminated. For the second stage of lexicon reduction, the lexicon is further reduced by comparing character strings using the key characters, with lexicon entries. For each of the key characters in the character strings, it is determined whether there is a mismatch between the key character and characters in a corresponding search range in the lexicon entry. If the number of mismatches for all of the key characters in a search string is greater than (1+(the number of key characters in the search string/4)), then the lexicon entry is eliminated. Accordingly, the invention advantageously accomplishes lexicon reduction, thereby decreasing the time required to recognize a line of cursive text, without reducing accuracy.

Claims

1. In a method for identifying key characters in cursive text, performing tasks comprising:

identifying a sequence of graphemes (G=(GR₁, GR₂, . . . GR_n)) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[1], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein iε

{1, . . . , n}, jε

{1, . . . , m}, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij for each segment (S_ij) for which m(S_ij) is ascertained, wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100}.

2. The method of claim 1 further comprising determining if the character corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained is any character in a specified group of characters.

3. The method of claim 2 wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

4. The method of claim 1 further comprising:

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

5. A method for identifying key characters in cursive text, comprising:

identifying a sequence of graphemes (G=(GR₁, GR₂, . . . GR_n) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[1], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein iε

{1, . . . , n}, jε

{1,, . . . , m }, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij, for each segment (S_ij) for which m(S_ij) is ascertained, wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100};

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

6. The method of claim 5 wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

7. An article of manufacture comprising a data storage medium tangibly embodying a program of machine-readable instructions executable by a digital processing apparatus to perform, in a method for identifying key characters in cursive text, tasks comprising:

identifying a sequence of graphemes (G=(GR₁, GR₂, . . . GR_n)) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[1], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein iε

{1,. . . , n}, jε

{1, . . . , m}, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij, for each segment (S_ij) for which m(S_ij) is ascertained, wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100}.

8. The article of manufacture of claim 7, the tasks further comprising determining if the character corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained is any character in a specified group of characters.

9. The article of manufacture of claim 8 wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

10. The article of manufacture in claim 7, the tasks further comprising:

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

11. An article of manufacture comprising a data storage medium tangibly embodying a program of machine-readable instructions executable by a digital processing apparatus to perform, in a method for identifying key characters in cursive text, tasks comprising:

identifying a sequence of graphemes (G=(GR₁, GR₂, . . . GR_n)) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[l], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein iε

{1, . . . , n}, jε

{1, . . . , m}, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij, for each segment (S_ij) for which m(S_ij) is ascertained, wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100};

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

12. The article of manufacture of claim 11, wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

13. A digital data processing apparatus, program to perform, in a method for identifying key characters in cursive text, tasks comprising:

identifying a sequence of graphemes (G=(GR₁, GR₂, . . . GR_n)) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[l], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein iε

{1, . . . , n}, jε

{1, . . . , m}, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij, for each segment (S_ij) for which m(S_ij) is ascertained, wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100}.

14. The apparatus of claim 13, the tasks further comprising determining if the character corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained is any character in a specified group of characters.

15. The apparatus of claim 14 wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

16. The apparatus of claim 13, the tasks further comprising:

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

17. A digital data processing apparatus programed to perform a method for identifying key characters in cursive text, the method comprising:

identifying a sequence of graphemes (G=(GR₁, GR_{2, . . .} GR_n)) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[l], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein iε

{1, . . . , n}, jε

{1, . . . , m}, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij, for each segment (S_ij) for which m(S_ij) is ascertained wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100};

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

18. The apparatus of claim 17 wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

19. An apparatus for identifying cursive text on an object, comprising:

an optical scanner;

a digital data processing apparatus communicatively coupled to means for detecting the cursive text on the object, the digital data processing apparatus being programmed to perform a method for identifying key characters in cursive text, the method comprising;

identifying a sequence of graphemes (G=(GR₁, GR₂, . . . GR_n)) in an image;

identifying in the sequence of graphemes, a plurality of segments S_ij=(GR_i[l], GR_i+1, . . . GR<sub>i+j−

1</sub>, wherein is iε

{1, . . . , n}, jε

{1, . . . , m}, and i+j−

1≦

n;

calculating, for each of a plurality of characters, a level of confidence ch_x(S_ij) that at least one of the segments (S_ij) represents a particular character of the plurality of characters;

ascertaining the maximum value m(S_ij) of ch_x(S_ij) for the plurality of characters for at least one of the segments (S_ij);

determining if m(S_ij) is greater than a threshold T for each segment (S_ij) for which m(S_ij) is ascertained; and

determining if m(S_ij)>

m(S_kl)+d, ∀

S_kl;

S_ij∩

S_kl≠

0, and S_kl≠

S_ij, for each segment (S_ij) for which m(S_ij) is ascertained, wherein S_kl comprises an environmental measure of said segment S_ij and dε

{1, . . . , 100};

ascertaining a feature value v of a geometric property i of the character class x corresponding with m(S_ij) for at least one segment (S_ij) for which m(S_ij) is ascertained;

determining an estimated density function d_ix of the geometric property i of the character class x corresponding with m(S_ij) for each segment (S_ij) for which the feature value v is ascertained;

calculating the geometrical confidence

20. The apparatus of claim 19 wherein the specified group of characters is “

l”

, “

1”

, “

r”

, “

o”

, “

t”

, “

f”

, and “

2”

.

21. The apparatus of claim 19, further comprising an automatic object feeder positioned to advance objects past the optical scanner.