JPS60189070A - Character input device - Google Patents

Abstract

PURPOSE:To obtain a correct decision result and perform easily input operation without any sensing of incompatibility by inputting only characters which have partial features on handwriting basis, and limiting the ranges of candidates for the input characters greatly. CONSTITUTION:The character input device processes character data electronically by using a speech recognition part 2 and a character recognition part 8 in combination, and when a handwritten character is recognized, partial features of the character is handwritten through a character radical feature pattern dictionary 21. Then, its reading is recognized and the word/phrase vocabulary belonging to the reading is selected in a word/phrase vocabulary dictionary 23 through a word/phrase vocabulary generation part 22. Then, this is compared by a character decision part 12 with speed recognition candidates obtained by the speech recognition part. Namely, only characters having partial features are inputted on handwriting basis and the ranges of candidates for the input characters are limited greatly to obtain correct decision results and also perform input operation easily without any sensing of incompatibility.

Description

DETAILED DESCRIPTION OF THE INVENTION [al] Technical Field of the Invention The present invention relates to a character input device that uses both vocal character input and handwritten character input in order to process characters in a data processing device.

[b) Background of the technology In recent years, voice recognition technology and character pattern recognition technology have evolved due to the development of data processing technology and the need for complex, sophisticated logic circuits and high-speed, large-capacity methods for algorithms such as voice recognition and character pattern recognition. This has been made possible by the realization of smaller, lower-cost highly integrated circuit devices (LSI) due to advances in semiconductor integration technology, and it has become possible to easily process data using these highly integrated circuit devices. Speech recognition devices that recognize unknown input speech and human-powered text devices that recognize patterns of two handwritten characters and convert them into text readings have been put into practical use.

In particular, we recognize unknown human voice and create monosyllabic character strings.
A voice recognition device has the function of converting this monosyllabic character string into multiple words/41 words by comparing each word/7JJ imperial dictionary, and 5 recognizes the character shape pattern of handwritten characters to read the characters. Handwritten character input devices and the like have been realized that have the function of creating a handwritten character and converting it into a plurality of words/idioms by comparing each word/idiom with the SK party dictionary based on the character pronunciation. However, recognition errors cannot be avoided in speech recognition and character recognition. However, since there is no correlation between the tendency of misrecognition in both cases, if we judge the results comprehensively from the results of both cases,
The ``Character Input Device'', which was filed on October 25, 1980, was thought to enable extremely high recognition rates and high-speed character input, and was developed by combining this voice recognition device and handwritten character input device into 41 The present invention is a proposal that improves the 81 LJ rate and character input using the 81 LJ rate of characters.

This is an improvement proposal for the ``Character Input Device'' filed on October 25, 1988.

(C) Prior Art and Problems Below, we will explain the earlier application filed on October 25, 1980 (= J-key single-character input device). Figure 1 shows the configuration of a conventional character input device. Show the diagram. 1 is a microphone, 2 is a speech recognition unit, 3 is a speech feature parameter dictionary, 4 is a kana-kanji conversion unit, 5 is a kana-kanji conversion dictionary, 6
1 is a tablet, 7 is an indicator pen, 8 is a character recognition unit, 9 is a character feature pattern dictionary, 10 is a reading conversion unit, 11 is a reading conversion dictionary, 12 is a character determination unit, 13 is a character/word dictionary, 1
4B Main control unit.

15 is a display control unit, and 16 is a display.

In FIG. 1, first, the procedure from inputting an unknown human voice through the microphone 1 to recognizing the voice will be explained. An unknown human voice input through the microphone 1 is converted into an electrical signal and input to the voice recognition unit 2 as a voice human signal. For example, if the voice recognition unit 2 inputs a voice human horn signal 9,
The pitch frequency, formant frequency, spectrum, linear prediction coefficient, LSI partial autocorrelation coefficient, etc. are extracted as feature parameters, and the extracted human speech feature parameters are combined with the speech standard feature parameters registered in advance in the speech feature parameter dictionary @3. They are compared and matched, and those with a high degree of similarity are selected as monosyllable recognition candidates. Next, the monosyllabic recognition candidates obtained by the speech recognition unit 2 are sequentially input to the kana-kanji conversion unit 4, and the kana-kanji conversion unit 4 creates a plurality of monosyllabic character string candidates from the sequentially input monosyllabic recognition candidates. Then, referring to the kana-kanji conversion dictionary 5, the character string codes are converted into a plurality of character string codes of kana-kanji.

Next, a procedure for recognizing handwritten characters input to the tablet 6 will be explained. The same characters as the human voice are handwritten on the tablet 6 using the pointing pen 7, and the coordinate data of the moving pointing pen 7 is sequentially read by electromagnetic detection or ultrasonic position detection, and the coordinate data is transferred from the tablet 6 to the character recognition unit 8.
I'm going to enter.

Coordinate data related to handwritten characters input to the character recognition unit 8 include 1, for example, stroke features 1 topological features,
Character feature parameters are extracted by pattern analysis of core line direction features, background part features, etc., and the extracted character feature parameters are compared with each standard character feature paran~ri in the character feature pattern dictionary 9 registered in advance. Then, the "reading" candidates to which the standard character feature parameters that are close to each other belong are selected and sent to the reading conversion unit 10. The reading conversion section 10 converts the "reading" candidates obtained by the character recognition section 8 into reading character codes of one handwritten character by referring to a reading conversion dictionary 11 prepared in advance.

With this, the input voice and handwritten character recognition procedure is completed, and each recognition candidate is input to the single character determination section 12. The character determination unit 12 compares and collates both the plurality of character string codes of the voice recognition results manually inputted by the kana-kanji conversion unit 4 and the reading character codes of the handwritten character recognition results inputted from the reading conversion unit 10, and identifies the same characters. Selecting a coat as a judgment result, further selecting a character/word code from a character/word dictionary 13 in which the character code of the judgment result is registered in advance as a meaningful character or character string code as a recognition result, The selected character/A'L word code is transmitted to the main control section 14 as a recognition result. The character/word code transmitted to the main control section 14 is sent to the display control section 15.

The display control unit 15 converts the character/word code into kana/kanji and displays it on the display screen of the display 16.

This is the conventional voice recognition function for unknown input voice.

This is a character input device that also has a function to recognize handwritten characters.

Next, to explain with an actual example, let us say that the characters that we want to judge as 1, for example, 9, are 1 connection. The voice input to microphone 1 is "Re-n-ra-ku".

To input characters by hand using instruction ben 7 on tablets 1-6,
"Contact". In this case, the character string candidates created by the kana-kanji conversion unit 4 as monosyllabic character string candidates from the monosyllabic candidates recognized by the speech recognition unit 2 are “E・N・Ra・＜”, “Re・
There are multiple words such as ``n・ra・＜'', ``re・n・ya・ku'', etc., and the voice recognition converted by the kana-kanji converter 4 may misrecognize it as ``Dengaku'', for example. There is. On the other hand, the 9-character recognition section 8 selects a "reading" candidate, and the reading conversion section 10 converts it into a character code as a candidate based on 1-character recognition.

For example, there are "contact", "contact", "contact", etc.

Accurate recognition is difficult even with character recognition alone, so the 5-character judgment unit 12 eliminates "Dengaku", "Tsunagi", etc., and inputs characters that correctly recognize "Tsunagi", a candidate for which both parties have selected the same character. It is a device.

The conventional character input device described above uses recognition result candidates based on an uncorrelated method of two handwritten character recognition in order to obtain a single determination result from multiple candidates obtained as a result of voice recognition Φ, and makes a comprehensive judgment. By doing so, an extremely high recognition rate can be obtained. However, although it is a normal practice for people to read while writing, there are many people who feel reluctant to write kanji correctly.

Therefore, this conventional method has the drawback that it takes time and effort to handwrite all the characters, which reduces efficiency.

In addition, the input operation for inputting two handwritten characters is troublesome and may cause a sense of discomfort and failure.

fd) Purpose of the Invention The present invention aims to improve the drawbacks of the conventional method of handwriting all characters, which is inefficient and can cause a user to feel uncomfortable and troublesome, and to provide a method that is easy to input without causing any discomfort. The purpose is

fe) Structure of the Invention The above object is to provide a character input device used in a processing device that electronically processes 9 character data, which includes a voice hook means for inputting unknown input voice and a character input device for manually inputting partial characteristic shapes of 5 characters. A monosyllabic character string is generated by comparing input voice characteristic parameters obtained by acoustically analyzing the voice information inputted by the voice input means with monosyllabic standard parameters registered in advance in a dictionary. a speech recognition unit that selects recognition candidates; and a first selection that selects a plurality of kana-kanji candidates by comparing the monosyllabic character string recognition candidates obtained by the speech recognition unit with a kana-kanji dictionary registered in advance in a dictionary. 1. Compare the partial input characteristic parameters obtained by pattern analysis of the partial characteristic character information inputted by the handwritten character input means with the partial characteristic character standard parameters registered in advance in the dictionary. A handwritten character recognition unit that selects a character “reading” candidate and a partial “reading” candidate obtained by the handwritten character recognition unit select words/idiom words belonging to the pre-registered partial character “reading”. Partial character “reading” from the dictionary
a second selection means for selecting a plurality of word/idiom candidates to which the candidate belongs; a plurality of kana-kanji candidates selected by the first selection means and a plurality of words/idioms selected by the second selection means; a determination unit that compares and determines candidates; the voice recognition unit; the handwritten character recognition unit; and the first selection unit.

said second selection means and a control section that controls said determination section, said plurality of kana-kanji candidates selected by said first selection means and said plurality of kana-kanji candidates selected by said second selection means. Enter the word/idiom candidate into the judgment section,
This is achieved according to the present invention, wherein the determination unit compares and determines the same candidate from among a plurality of candidates under the control of the control unit.

In other words, the voice recognition unit inputs unknown input voice and recognizes the voice, and the character recognition unit inputs 9 handwritten characters and recognizes the characters.
5 moves in a character input device used in a processing device that electronically processes character data using a computer? : Partial character recognition in case of character recognition! Write the +, 'l signs.

``Reading'' of partial characteristics of characters in literary magazine 8ii1i L
, a word/tu 1-word dictionary that is equivalent to a single imperial dictionary consisting of radicals that are partial partial readings of characters, and a Kanwa dictionary that is a dictionary of idioms for each kanji character. picture.

Select JR4 words/idiom words to which the 2nd partial ``yomi'' belongs from the word/idiom dictionary that belongs to the 2nd partial ``yomi'', and in the 1st judgment section select the speech recognition candidates that are the same as the speech recognition candidates obtained by the speech recognition section. By configuring the system to compare and judge candidates, only characters with one partial characteristic are manually handwritten, and the range of candidates is greatly limited by the input character of the radical or kanji-character. The correct judgment result is changed by comprehensively judging human characters or input character string candidates obtained by handwritten input character recognition of a part of a character or a part of a character string, and a plurality of character string candidates by voice recognition.

An object of the present invention is to provide a character input device that can easily perform input operations while ensuring a high recognition rate.

Embodiment of the Invention An embodiment of the present invention will be described below. FIG. 2 shows a block diagram of the structure of a character input device according to the present invention, and the same objects are not shown with the same reference numerals as in FIG. 1.

Reference numeral 21 indicates a character radical feature pattern dictionary, 22 indicates an idiom/idiom quantity creation unit, and 23 indicates a word/idiom dictionary. The character radical feature pattern dictionary 21 is a dictionary in which feature parameters of radical parts of characters are registered in advance as standard character radical parameters.
]J, r thread”...marks? 1 (registered for each reading corresponding to the character radical feature parameter. The word/idiom creation unit 22 converts one word/tzu belonging to the “reading” candidate obtained by the character recognition unit 8 into a word/41 words. Select from dictionary 23
It creates character string candidates. In the word/idiom quantity dictionary 23, a plurality of words/idioms belonging to each radical of a character are registered in advance for each radical quantity.

In FIG. 2, the steps from the voice recognition unit 2 recognizing the human voice input through the microphone 1 to inputting multiple character string codes from the kana-kanji conversion unit 4 to the character determination unit 12 are as described above. Since it is the same as , it is omitted. By handwriting the same radical of a character as the input voice with an instruction pen 7 on a tablet 6, the coordinate data is inputted to a character recognition unit 8.

The character recognition unit 8 extracts the character radical feature parameters of the coordinate data, and compares the extracted character radical feature parameters with the standard character radical feature parameters of the character radical feature pattern dictionary 21 registered in advance. After comparison and collation, the "reading" candidate to which the closest standard character radical feature parameter belongs is selected and input to the i11 word/QpH imperial party creation section 22. The word/idiom creation section 22 reads out a plurality of if=word/idiom candidates that belong to the "reading" candidates obtained by the character recognition section 8 from the word/idiom quantity dictionary 23 and manually inputs them to the δftL7j conversion section 10. . A plurality of word/idiom candidates inputted to the J/SO/Lomi conversion section 10 are converted into pronunciation character codes of handrail characters with reference to the reading conversion dictionary 11, and then converted to the reading character code of the handrail characters.
Enter. Regarding both the plurality of character string codes of the voice recognition results from the kana-kanji conversion section 4 input to the character determination section 12 and the plurality of character string codes 1 of the character recognition results of handwritten character radicals from the reading conversion section 10 After comparison and collation, the same recognition candidate characters are displayed on the display 16 as recognition results in the same procedure as described above.

Next, to explain with a practical example, 5, for example, 1 as in the previous description.
When inputting the word ``contact'' manually, the voice recognition side
Say "N-ra-ku" and say "Contact."

It is the same as the previous description that a character string code such as "Dengaku" is input to the character determination section 12, but on the single character recognition side, the tablet 6 is sent one of "Contact" to "Car". It is good to handrail the radical of ``ren''.

The character recognition section 8 recognizes the coordinate data "car" from the tablet 6 and inputs the pronunciation of "car" to the word/9JH edict light creation section 22. Vocabulary/idiom edict? In the creation section 22, jQ
The various quantities with a "car" bias such as "trajectory", "contact", "renju", "light", etc. registered in the word/idiom word quantity book 23 are read out and input to the reading conversion unit 10. and converted to a two-letter code. The character determination unit 12 selects "Dengaku", "Kyoku", "
``Renju'', ``Kikai'', etc. are divided by II, and ``Contact'' included in both candidates is recognized and determined and displayed on the display 16.

2For example, if you enter the word "recognition",
Say ``ni-n-shi-ki'' and write the ``word''. Candidates based on voice recognition include "recognition", "shinshiki", etc., and word candidates that include kanji with "go" as a radical, "gojo J, r; ti,""correction","Recognition", etc. are selected. As both of the candidates are included, "recognition" is the determination result and is displayed on the display 16.

The above embodiment of the present invention has been explained using a dictionary in which standard radical feature patterns for two characters are registered in the character radical feature pattern dictionary 21. It goes without saying that the same result can be obtained even if a kanji-character-based word/idiom vocabulary dictionary is registered in the Ikugodo dictionary 23. The selection can be made depending on whether it is a part of a character or a part of a character string.

+gl As described in detail of the invention, a voice recognition unit that inputs unknown input speech and recognizes the voice, and a character recognition unit that inputs handwritten characters and recognizes the characters (processing of electronically processing character data using jf). In the character input device used in the device, 5 handwritten character recognition is performed, ``ζ 1 character partial feature is written, 1 character partial feature 1 reading'' is character recognized, and 1 radical is the bias of the character belonging to the partial ``reading''. It is equipped with a word/idiom quantity dictionary equivalent to the Kanji dictionary, which is a dictionary with a word quantity dictionary and a kanji/idiom quantity dictionary for each character. The present invention is configured to select the word/Qp5 word spirit candidates to which "yomi" belongs, and compare and determine the same candidates as the speech recognition candidates obtained by the speech recognition section in the first judgment section. By handwriting the name of the character you have.

The range of input character candidates is greatly limited, and correct judgment results can be obtained by comprehensively judging the input character or input character string candidates obtained by handwritten character recognition and the multiple character string candidates obtained by voice recognition. , the input operation becomes easier, and
This has the effect of ensuring a high recognition rate.

[Brief Description of the Drawings] Figure 1 shows a block diagram of a conventional character input device.
FIG. 2 shows a block diagram of a character input device according to the present invention. In the drawing, 1 is a microphone, 2 is a voice recognition unit, and 3 is a microphone.
is a voice feature parameter dictionary, 4 is a kana-kanji conversion unit, 5 is a kana-kanji conversion dictionary, 6 is a tablet, 7 is an indicator pen, 8
is a character recognition unit, 10 is a reading conversion unit, 11 is a reading conversion dictionary, 12 is a character judgment unit, 13 is a character/word dictionary, 1
4 is a main control section, and 15 is a display control section. Reference numeral 16 indicates a display, 21 a character radical feature pattern dictionary, 22 a word/twenty-word quantity creation unit, and 23 a word/idiom quantity dictionary.

Claims (1)

[Scope of Claims] A character input device used in a processing device that electronically processes character data, comprising a voice input means for inputting unknown input voice and a handwriting device for handwriting a partial characteristic shape of one character. The input voice feature parameters obtained by acoustically analyzing the voice information input by the character input means and the voice input means are compared with monosyllabic standard parameters registered in advance in a dictionary to identify monosyllabic character string recognition candidates. With a voice recognition unit to choose from. a first selection means for selecting a plurality of kana-kanji candidates by comparing the monosyllabic character string recognition candidates obtained by the speech recognition unit with a kana-kanji dictionary registered in advance in a dictionary; The partial character input characteristic parameters obtained by pattern analysis of the partial characteristic character information inputted by the handwritten character input means are compared with the partial characteristic character standard parameters registered in advance in the dictionary, and one partial character " A handwritten character recognition unit that selects ``@mi'' candidates. Words/words belonging to the pre-registered partial character "yomi" are determined by the partial "yomi" candidates obtained by the handwritten character recognition unit.
a second selection means for selecting a plurality of word/idiom candidates to which the partial character "reading" candidate belongs from the idiom dictionary; and a plurality of kana-kanji candidates selected by the first selection means and the second selection means. a determination unit that compares and determines a plurality of word/idiom candidates selected by the selection means, the speech recognition unit, the handwritten character recognition unit, the first selection unit, the second selection unit, and the determination unit and a control unit that controls the plurality of kana-kanji candidates selected by the first selection means and the plurality of word/idiom candidates selected by the second selection means to the determination unit. A character input device characterized in that the character input device is configured such that the determination unit compares and determines the same candidate from among a plurality of candidates under the control of the control unit.