-
公开(公告)号：JPH06348904A
公开(公告)日：1994-12-22
申请号：JP5923994
申请日：1994-03-29
Applicant: IBM
Inventor： TETSUNOSUKE FUJISAKI , UIRIAMU DEIBITSUDO MODORIN , KOTSUTAPARAMU EMU MOHIYUUDEIN , TAKAHASHI HIROYASU
Abstract: PURPOSE: To improve the error rate of a handwritten character recognizing process. CONSTITUTION: A method for recognizing handwritten character includes (a) steps A and B for giving time-sequentially arranged dynamic stroke information in response to a handwritten character input from a user, a step C for converting (b) the time-sequentially arranged dynamic stroke information and (c) time-sequentially arranged stroke information into static stroke information, (d) a step D1 for deciding a second list containing at least one possible character indicated by the static stroke information, (e) and steps E and F for combining first and second lists for giving a third uniformized list containing at least one element indicating the most possible character the time-sequentially arranged dynamic stroke information intends to indicate.
-
公开(公告)号：JPH0684006A
公开(公告)日：1994-03-25
申请号：JP4319393
申请日：1993-02-08
Applicant: IBM
Inventor： HOMAYOON SADOA MOHAMEDO BEIGI , TETSUNOSUKE FUJISAKI , UIRIAMU DEIBITSUDO MODORIN , KENESU SUTEIIBUN UENSUTORATSUP
Abstract: PURPOSE: To surely avoid an error by utilizing a dictionary for postprocessing error correction in online recognition of hand-written character. CONSTITUTION: In a verification block 26, a recognized character string is collated with the dictionary in a block 28 to check whether the dictionary has a word having this spelling or not. If the dictionary has this word, the word is given to a cache integrated block 30, and the output word is given to a line 31. If the dictionary doesn't have the word, character collation marks are calculated in a block 32; and if marks per character reach a determined level, the word is outputted to the block 30, and the output word is given to the line 31, and stroke collation is calculated in a block 34, and a hypothesis of best match is marked and preserved together. The final word hypothesis of best collation marks is inserted to a global word hypothesis list, and thereafter, the output of a collation block 34 is given to a spelling support block 36, and words resembling the character string are searched in the dictionary and are inserted to the global word hypothesis list.
-
Patent Agency Ranking3.发明专利
公开(公告)号：JPH0620099A
公开(公告)日：1994-01-28
申请号：JP22481291
申请日：1991-08-09
Applicant: IBM
Inventor： TOOMASU EDOWAADO CHIEFUARASU , TETSUNOSUKE FUJISAKI , JIYOONKI KIMU , CHIYAARUZU KAASON TATSUPAATO
Abstract: PURPOSE: To establish a character prototype set having a sufficient applicable range, proper expression, and mutual sufficient distance. CONSTITUTION: This method includes a stage for inputting many strokes being the constituting strokes of a mark to be recognized by this system, stage for forming a first set (CHR) constituted of more than one marks from the inputted strokes, stage for meaning the more than one characteristics of the similar configuration mark being the element of the first set for forming a second set (AVG) constituted of a prototype mark to be compared with the set of the inputted strokes during the usage of the system, stage for resolving the prototype marks of the second set into the constituting strokes, and stage for forming a third set (STK) constituted of the constituting strokes of the second set to be compared with the inputted strokes during the usage of the system. Also, this method includes a stage for identifying the mark of the second set corresponding only to the single mark of the first set, and a stage for deleting the identified mark from the second set.
-
公开(公告)号：JPH05217025A
公开(公告)日：1993-08-27
申请号：JP22042092
申请日：1992-08-19
Applicant: IBM
Inventor： TETSUNOSUKE FUJISAKI , JIYOONKI KIMU , JIYOOJI JIYOZEFU RIIBUMAN , CHIYAARUSU KAASON TAPAATO
Abstract: PURPOSE: To reduce similarity between different prototypes in an on-line handwritting recognizing system by interactively removing similar prototypes to be frequently confused by a user. CONSTITUTION: A box is touched by a pen tip 14, and when the selection is executed to prepare a new prototype (n), a new window is opened on a display screen and a label (n) newly intended by a user is handwritten. A prototype processor 24 prepares the new prototyp (n), which is added to a group of prototypes and stored in a character prototype storage device 20. In the case of storing a new handwritten character for the intended label (n), a new prototype is prepared from the intended label (n) of the new hand written character by the processor 24 and added to the contents of the device 20. In the case of canceling the new handwritten character of the intended label (n), the processor 24 removes the handwritten character from the device 20.
-
公开(公告)号：JPH0962793A
公开(公告)日：1997-03-07
申请号：JP12955496
申请日：1996-05-24
Applicant: IBM
Inventor： TETSUNOSUKE FUJISAKI , SABITA SHIYURINIBASAN
Abstract: PROBLEM TO BE SOLVED: To provide system and method for executing an increment retrieval in a confused matrix coding dictionary by using the character string that a recognizing device returns to predict the word of the handwritten input field of an application program. SOLUTION: In these system and method, a handwritten character is inputted in the data input field of an application program. The handwritten character is recognized and a cluster code is assigned. A character string is featured by the cluster code of each character. The cluster code of the character string is compared with the character string of the cluster code showing the word in a dictionary. All or some of equal words are displayed for a user and the selection from the displays is executed.
-
公开(公告)号：JPH0684016A
公开(公告)日：1994-03-25
申请号：JP1634593
申请日：1993-02-03
Applicant: IBM
Abstract: PURPOSE: To resolve problems apt to bring about confusion or case grammar by determining whether a mojority of characters in a word is numeric or alphabetic characters and determining whether a majority of alphabets is upper-case or lower-case in the case of alphabetic characters and determining whether characters should be changed to similar prototype characters in accordance with the case class of the majority. CONSTITUTION: In a block 20, it is determined whether a character is in the case class of alphabetic characters or the case class of numeric characters, and it is determined whether the character is in the case class of upper-case characters or the case class of lower-case characters. With respect to the whole of an output word, the number of members free from confusion of cases in each case class is counted, and a case is determined in accordance with majority of numbers of members free from confusion of cases. Thereafter, case grammar is applied if a character of the word is a prototype character apt to bring about confusion of cases, and the character is converted to a corresponding equivalent prototype character in accordance with the case group of the majority. For example, 'z' and '1' of '49zz18' are converted to '2' and '1' respectively.
-
公开(公告)号：JPH04296989A
公开(公告)日：1992-10-21
申请号：JP41665790
申请日：1990-12-29
Applicant: IBM
Inventor： TETSUNOSUKE FUJISAKI
Abstract: PURPOSE: To provide a method and device for recognizing characters for applying the recognition of the constraint condition driving of an on-line document character mark. CONSTITUTION: At the time of identifying an effective mark from a series of handwritten strokes, this method includes a step for generating plural stroke labels having an evaluation point combined with each other corresponding to the handwritten strokes, step for processing those strokes labels by a beam retrieval technique, and identifying the stroke label indicating the effective mark, and step for identifying the characteristic of the effective mark by combing those identified stroke labels. Here, a constraint condition filter 36 can connect each serial filter so as to be freely switched like a chain, and those filters function by connecting or separating an input with or from the stroke labels, and applying a path bypassing a filter block. An application programmer can use the plural constraint condition filters 36. Also, the programmer specifies which constraint condition filter is applied to a specific stroke row.
-
-
-
-
-
-
Search Results
7
records
(took 0.017 seconds)
