CN1210301A - Phonetical vein inputting method and phonetical letter association method - Google Patents
Phonetical vein inputting method and phonetical letter association method Download PDFInfo
- Publication number
- CN1210301A CN1210301A CN 98114665 CN98114665A CN1210301A CN 1210301 A CN1210301 A CN 1210301A CN 98114665 CN98114665 CN 98114665 CN 98114665 A CN98114665 A CN 98114665A CN 1210301 A CN1210301 A CN 1210301A
- Authority
- CN
- China
- Prior art keywords
- phonetic
- word
- speech
- user
- confirmed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
A Chinese-character sound ripple input method and phonetic association method is a speech input method for computer. The sound ripple method features that the sound ripple characteristics of user speech are compared with prestored sound ripple characteristics in computer to determine the read-in contents while give out correspondent phonetic letters and homonyms, from which user can choose one. The phonetic association method features that the phonetic letter group provided by the sound ripple method is compared with that in phonetic phrase library to find out relative phonetic letters and directly determine its words.
Description
The present invention relates to a kind of with the phonetic entry Chinese character to computer method and a kind of method of utilizing association to accelerate Chinese character input speed.
Since computing machine entered China, the Chinese character input just was a difficult problem always.Chinese character input method of today mostly is the keyboard input greatly, and method is not hundreds of down, but real practicality, approximately also with regard to tens kinds.Wherein, more representative with region-position code method, phonetic method, the Five-stroke Method method.The shortcoming of region-position code method is that the user is difficult to remember numerous codings; The phonetic rule requires the user to be familiar with phonetic, and this is difficult for grasping for the people who is accustomed to the use of accent; The Five-stroke Method is present the widest popular a kind of method, and by numerous professional persons are adopted, but for amateurish personage, especially the not good people of memory will learn more than 100 radical and the various group word rule of divining by means of characters by heart, neither a nothing the matter.Generally speaking, above method all fails to break away from the constraint of Chinese character " font complexity ".
Understand language if calculate function, be undoubtedly method to computing machine with the voice input characters.In the nineties, the speech recognition technology of computing machine has had very big development, as having put down in writing " trans-oceanic automatic translation phone " test of being undertaken by Japan and the United States, moral 3 states at the beginning of 93 in " computer funny remarks " (literary composition remittance in March, the 95 version Ye Yonglie of publishing house, Ling Qiyu work), 3 sides say is separately language, has realized instant translation accurately basically; And the vocal print credit card that U.S. Bel's communication research institute releases, but can determine one's identity: need say when the user gets neocaine that a password is as keeping on file by sound, when using card, to say password, carry out voiceprint analysis, thereby confirm user's identity by computing machine; More there is a tame u s company to produce the acoustically controlled computerized typewriter that the both hands handicapped person uses.
The speech recognition of computing machine is that digitized voice data made in people's voice, compares, judges with sample.Traditional speech recognition technology is undertaken (seeing " multimedia " by " phonetic entry → acoustic treatment → word boundary detection → feature vector extraction → word pattern coupling → identification output ", publishing house of the National University of Defense technology, publish in January, 1996), system's recognition result in the noiselessness environment is good, but the hurried decline of performance is being arranged under the noise situation.The new system that releases has increased noise immunological learning function, thereby accuracy of identification increases.Two kinds of systems generally all are used for unspecified person.Because be used for specific people's system, its usable range will be subjected to significant limitation.
In the application aspect the Chinese character input, the system of Tui Chuing is quite a few in recent years as for speech recognition technology.But up to now, do not popularize with the method for phonetic entry Chinese character is a large amount of, the key here is that computing machine is enough not high yet to the accuracy of speech recognition.Reason is that everyone sound characteristic is all inequality, adds that China is very large in area and thickly populated, and each local accent is all different, speaks even the also difficult total number of Chinese itself is understood various accent, in addition is that fuzzy ability can not show a candle to human computing machine.In addition, noise also is the key factor that influences the computer speech recognition capability.Therefore, allow the voice of computer Recognition unspecified person have any problem, can influence the accuracy of identification, the application that has also therefore influenced the computer speech input is with universal.
The object of the present invention is to provide a kind of method, give full play to the advantage of Chinese, and utilize the characteristic and the reappearance of human vocal print, improve the accuracy of computer speech identification with phonetic entry Chinese character (giving computing machine).
Modern science points out that people's sound has characteristic and the reappearance similar with fingerprint, thereby also is called vocal print.Same individual reads same word, and the reappearance of its vocal print is extraordinary.Therefore, allow the specific people's of computer Recognition voice, the accuracy of its identification be high more than the identification unspecified person, and first method of the present invention---computer sound and grain input method (be called for short the vocal print method, down with) designs just based on this.
The general plotting of vocal print method is: allow computing machine remember whole Chinese everyday character phonetic features of specific user earlier, after, when this user will want that the Chinese character of importing is read to computing machine, the phonetic feature that computing machine is about to read in compares with the sample of remembeing in advance, thereby determines to read in the word content of voice.
According to totally 1332 of the Chinese character pronunciations of " modern Chinese dictionary " (Commercial Press in Dec, 1978 publish) record, consider seldom usefulness of a malapropism sound, can think that the pronunciation of Chinese everyday character (has comprised tone) about 1300.Each word sound can have a plurality of phonetically similar words (at least one)." understand " this 1,300 sounds as long as calculate function, know pairing phonetic of each sound and phonetically similar word thereof, so, when the word that we will import is read to computing machine, computing machine just can be presented at phonetic and whole phonetically similar word of word sound on the display, selects for us.
At first need a sound processing apparatus, hereinafter referred to as " standard vocal print maker ", it can be with the word tone signal of being sent here by microphone, and word sound (single syllable) carries out standardization one by one, make digitized voice data (hereinafter referred to as " standard voiceprint sign indicating number "), deposit memory in.
Everyone must carry out " input of the standard that prestores vocal print sign indicating number " earlier before using the vocal print method for the first time.Order by computing machine is pressed dictionary will be presented on the display one by one from about 1,300 the phonetic of ā → zu ò, and points out with the most frequently used phonetically similar word.The user follows screen and clearly reads each word sound one by one.Sound is through microphone → standard vocal print maker, and each word change of tune becomes one group of standard voiceprint sign indicating number to deposit in the storer.For example, screen shows that " ā Ah ", person to be used run through that " Ah "'s sound, computing machine is receiving that " behind the standard voiceprint sign indicating number of Ah "'s sound, the next sound " á " with ā is presented on the screen again, goes down so one by one, till " zu ò does ".Like this, computing machine just remembered that the user read from 1,300 of ā → zu ò (about) the word sound.
Through after the step of " input of the standard that prestores vocal print sign indicating number ", the user has imported Chinese character with regard to available sounds.Way is: the word that will import (such as " in " word) read, the word sound is through microphone → standard vocal print maker, change into standard voiceprint sign indicating number, computing machine compares this group information code with 1,300 groups that prestore, therefrom find out the most similar person (because identical chance is little, so confirm as the most similar person identical), then the phonetic of this sound and phonetically similar word are presented on the screen (" in " demonstration of word sound is: among the zh ō ng＜loyalty, inner feelings, clock, eventually, handleless cup, Zhong, alarmed and panicky 〉), select (by " 0 " or " acknowledgement key " for the user, middle word promptly is selected, and remaining word and phonetic disappear).
As can be seen, after everyone only need carry out once " input of the standard that prestores vocal print sign indicating number ", just can repeat constantly on same computing machine, to use vocal print method input Chinese character, and computing machine is ignored other people sound.During practical application, as long as everyone exist oneself " standard voiceprint sign indicating number " in the floppy disk, just can be with floppy disk use vocal print method to any one computing machine that vocal print method function arranged.1,300 sounds are deposited people's computing machine, if one of per second takes 22 minutes; If 2 seconds one, need 43 minutes only.This paying of putting things right once and for all be said " value! ".
Owing to factors such as being subjected to neighbourhood noise is disturbed, also very difficult 100% repeat with the vocal print of word sound with the people.So in practice, and do not require the vocal print sign indicating number to repeat be 100%, and only require " similar ".As for similar degree, should see actual effect and decide, be principle with " the try one's best big detection probability and the little error rate of trying one's best ", therefrom find equilibrium point.Promptly when noise, other people voice are vetoed, user's voice are confirmed as far as possible.
" input of the standard that prestores vocal print sign indicating number " step should be carried out in the quiet environment of trying one's best.In case of necessity, can repeat one to twice, so that the vocal print of all Chinese-character pronunciations that allow computing machine accurately remember the user to be read.
The vocal print method does not also require that the user necessarily uses mandarin, and any dialect can both use, even habitually wrong pronunciation yet harmless (because computing machine is by repeated regular word selection).
Generally speaking, as long as understand the people of Chinese (Chinese), just can use the vocal print method.
Realize the vocal print method, at first will work out cover correspondent computer software---a sound and grain input method software.
Now illustrate the part operation situation of sound and grain input method software.For example to import " People's Republic of China (PRC) ", the user at first aim at microphone read " in " the word sound, graphoscope just shows
01201 " among the zh ō ng (loyal, eventually ...) " read " China " sound again, show and become " among the zh ō ng (loyal,
↑ 2012 eventually ...) h ú a draws (cunning, China) ", on keyboard,, be shown as " China ", again by " 2 "
↑ ↑ read " people " shows " Chinese r é n people (benevolence, the ninth of the ten Heavenly Stems, appoint) ", reads " people " again, is shown as
↑
0123012 " Chinese r é n people (benevolence, the ninth of the ten Heavenly Stems, appoint) the m ú n people (jade-like stone, Min ...) " read " be total to " again, and are apparent
↑
012301201 be shown that " the m í n people (jade-like stone, Min ...) g ò ng tribute (altogether, for Chinese r é n people (benevolence, the ninth of the ten Heavenly Stems, appoint)
↑ 23 Gong, Hong) ", by " 1 " key, show to change " the Chinese people altogether " into.Read again " with ", show
↑
012 be shown " the Chinese people altogether h é river (what, close ...) ", cursor " ↑ " is moved to right to " () "
↑
012345678 times, be shown as " the Chinese people altogether h é river (what, close, nuclear, lotus, box and, standing grain, jaw,
↑ 9 He ...) ", by " 6 " key, be shown as " Chinese people's republicanism ", read " state ", be shown as
↑
012 " Chinese people's republicanism gu ó state (Guo, popliteal ...) ", by " affirmation " key, be shown as " China
↑ people's republic ".
↑
Illustrate: symbol " ... " in " () ", expression still has phonetically similar word not show.General when cursor not in " () " following time, the phonetically similar word in " () " only shows 2 (also can consider 3 or 4), all the other usefulness " ... " expression, but, then all show (as top " people " and " tribute " word) as if 5 of the few mistakes of phonetically similar word in " () ".During the word selecting not show in " () ", cursor " ↑ " can be moved to below " () " interior first word, promptly showing 9 phonetically similar words in " () " (all shows when being less than 9,9 still have later, with " ... " expression, as above " river " word), if phonetically similar word is a lot, as long as " ↑ " toward moving to right one, promptly showed next group 9 phonetically similar word (last group of disappearance), because maximum phonetically similar words is 82 in " () ", deduct " () " outer, just 81, so when moving on to " 9 " number position when " ↑ ", the 82nd word of maximum 82 phonetically similar words also shows.Numeral above " () " interior word " 1,2 ... " the expression position, selective usefulness.By " 1～9 " key, the word of " () " interior relevant position of cursor place sound just is identified, and by " 0 " or " affirmation " key, the word (0 position) on " () " outer phonetic right side just is identified.The word that needs to confirm can not confirmed and direct read next word, when the word of back is identified when " 0 " position, the unacknowledged word in front can be identified also that (computing machine selects " 0 " position word to confirm automatically, as top " China ", " China " when word is identified, " in " word is identified automatically).When below " ↑ " is parked in room or confirmed word, can import the numeral (because not had the word that to confirm) of " 0～9 " by " 0～9 " key.During with vocal print method input Chinese character, the keyboard input is still effective, can use keyboard input digit, symbol, English alphabet simultaneously.
Through rough estimates, phonetically similar word accounts for 60.5% of sum (1300) at 6 with interior sound, so 60% word only can be selected by a key.Phonetically similar word accounts for 77.3% at 10 with interior sound, and phonetically similar word accounts for 94.1% at 19 with interior sound, so the word less than 6% that needs button could confirm more than 3 times.
When program editing, operation, also available vocal print method is gone input digit, various symbol, order etc., has so both made things convenient for the use of Chinese software, can quicken the Chinesizing of various softwares again.Such as symbol " (),＞,＜,=", can use respectively " left and right, large and small, etc. " representative of word sound, and for example order " IF " (if) representative of available " vacation " word sound, these represent the word sound at Chinese editor interval scale Chinese character sound, at program editing, running software interval scale symbol order, as long as handled just can not cause confusion.As for 26 English alphabets, and " 0～9 " ten numerals, should directly import with keyboard.
Chinese character is imported in the process of computing machine with voice, how speech conversion being become corresponding character is a key issue.The way that the vocal print method is adopted is: the word sound that the user is read changes into corresponding phonetic, provide the pairing whole phonetically similar words of this phonetic simultaneously, select (to be " sequence to be confirmed " for the user to call these phonetically similar words in the following text, it is the whole phonetically similar words by the phonetic of a word, become by the series arrangement of " commonly used row before ", first word with sequence to be confirmed is called " the most frequently used word " simultaneously).
The vocal print method can convert the word sound to corresponding phonetic, and this is a much progress, but phonetic also is not literal after all, each phonetic can be corresponding from a phonetically similar word that does not wait to dozens of, on average about six.Converting phonetic to required literal this step, if all finished by the user, is whole advantages of failing to bring into play Chinese character eventually.
Second method of the present invention---the purpose of phonetic association method, be to aim at the vocal print method a kind of association technology is provided, by the phonetic that is provided by the vocal print method is carried out association, find out vocabulary (and phrase) phonetic wherein, and confirm into corresponding character automatically by computing machine, thereby alleviate user's workload, and accelerate the speed of Chinese character input greatly.
One of characteristics of Chinese character are that phonetically similar word is numerous, when we hear a word sound, are easy to expect pairing several phonetically similar words commonly used of word sound, but are difficult to judge specifically be which word., (speech herein all refers to double-tone and the polyphonic word more than two sounds) often just can be write out speech when we hear a speech.That is to say that we can determine the literal of speech according to the pronunciation of speech.Though this is that homonym greatly reduces because the Chinese phonetically similar word is many, number does not have homonym mostly.The phonetic association method carries out association just on this basis.
The first step that realizes the phonetic association method is to set up pinyin lexicon.Pinyin lexicon is pressed certain regularly arranged composition by vocabulary, phrase (comprising phrase, idiom, Chinese idiom etc., down together) and their phonetic.Speech in the pinyin lexicon, phrase are from following several respects:
1, the whole little clauses and subclauses (comprising speech and phrase) in " modern Chinese dictionary " (78 years versions are hereinafter to be referred as " dictionary ") under the individual character entry, and the phrase of listing under the part individual character entry (referring to a large amount of phrases for explaining that individual character is listed in " dictionary ");
2, a large amount of new term new phrases of recent two decades appearance, the predicate of especially scientific and technological aspect;
3, software package can be made in professional, the speech of professional, phrase, by the user on demand optionally in the computing machine of input oneself;
4, movable dictionary, by user oneself import at any time need (and not having in the storehouse) between and phrase (also can delete at any time).
According to " dictionary " contained Chinese character pronunciation is 1332 altogether, and the corresponding phonetic of every sound is called individual character phonetic or single-tone phonetic (this phonetic comprises tone) at this, and the phonetic of speech and phrase is called speech phonetic and phrase phonetic.
The queueing discipline of speech, phrase and their phonetic is as follows in the pinyin lexicon:
1, be that clauses and subclauses are set up a little dictionary with each individual character phonetic, specifically certain sound can be called certain (phonetic) dictionary (for example " b á o dictionary ").To be included in the single-tone dictionary of this sound with all speech, phrase and the phonetic thereof of certain sound beginning.For example, b á o dictionary, speech, phrase and phonetic thereof with the beginning of b á o sound in " dictionary " have " hail (b á ozi) ", " pancake (b á o b ǐ ng) ", " crisp fritter (b á o cu ì) ", " thin plate (b á o b ǎ n) ", " thin slice (b á o pen) " etc., all are included in the b á o dictionary.
2, each the speech phonetic in the single-tone dictionary, phrase phonetic are all as little clauses and subclauses, and these little clauses and subclauses are arranged according to their lexicographical order, and promptly the order with English 26 letters is a preface, and each speech (phrase) phonetic is compared arrangement.Speech phonetic, phrase phonetic are made of single-tone phonetic because first single-tone phonetic is all identical, so contrast since first letter of second single-tone phonetic, first is alphabetical to contrast second when identical again, the rest may be inferred; Then contrast tone when the letter of second single-tone phonetic is all identical, tone is arranged by the order of " high and level tone, rising tone, go up sound, falling tone, softly "; Second single-tone phonetic contrasts the 3rd single-tone phonetic again when identical, the rest may be inferred.As the little clauses and subclauses ordering in the top said b á o dictionary be: [b á o b ǎ n] [b á o b ǐ ng] [b á o cu ì] [b á open] [b á ozi].
3, the speech in the single-tone dictionary, phrase (referring to literal) become only word content in the little clauses and subclauses of speech (phrase) phonetic immediately following after its corresponding phonetic.So just make computing machine after finding speech (phrase) phonetic, can provide corresponding literal immediately.The situation (overall picture) that top b á o dictionary is arranged is as follows:
[báo]
[b á o b ǎ n] thin plate
[b á o b ǐ ng] pancake
[b á o cu ì] crisp fritter
[b á o pen] thin slice
[b á ozi] hail
4, when two (and more than) speech (or phrase) when having identical phonetic, (in fact this is exactly homonym after they should being come simultaneously the little clauses and subclauses of its phonetic, such as " making an inventory of " and " refreshment " all return [ch á di ǎ n ］ in the little clauses and subclauses), and the series arrangement of press " row commonly used is preceding ".
5, the lexicographical order of 1332 single-tone dictionaries being pressed single-tone phonetic (clauses and subclauses) is arranged when identical (all alphabetical press the arrangement of tone order), promptly constitutes total pinyin lexicon.Arrangement in aligning method and the little clauses and subclauses is similar.
In pinyin lexicon, the arrangement of 1332 single-tone phonetic clauses and subclauses is in fact identical with " dictionary ", but the arrangement in each single-tone phonetic clauses and subclauses (being the single-tone dictionary) is then different with " dictionary ", and single-tone phonetic, speech, phrase and their phonetic are only arranged in the pinyin lexicon, and do not have in " dictionary " various explanations and individual character entry word, speech.
Now again for a complete single-tone dictionary example:
Example 1:[mi ǎ o]
[mi ǎ o bi ǎ o] stopwatch
[mi ǎ o ch ā j ù] parsec
[mi ǎ o m á ng ］ remote, stretch as far as the eye can eye
[mi ǎ o yu ò y ā n y ú n] is as vague as mist
[mi ǎ o sh ì] despises
[mi ǎ o w ú r é n j ì] is remote and unihabited
[mi ǎ o w ú sh ē ng x ī] vast silence
[mi ǎ o xi ǎ o] is negligible, tiny
[mi ǎ o zh ē n] second hand
Realizing second step of phonetic association method, is the software that design one cover can satisfy following affirmation process, is example with " we are eating refreshment " now, and the affirmation process is described:
What 1, the vocal print method adopted is the way of word for word importing, word for word confirming.But with the phonetic association method time, need change " input is sentence by sentence confirmed into literal automatically by the word sound that computing machine will be confirmed earlier, and remaining part unconfirmed is confirmed by the user again " into.
2, the user will＜we are eating refreshment (＜in word table show pronunciation, down together), word for word clearly read to computing machine, whole sentences and phrases are intact, computing machine with the result that the vocal print method draws be " my (...) m é n door (...) z à i of w ǒ 00000 (...) ch í eat (...) ch á examine (...) di ǎ n 0 point (...) " (for simplicity, the word in () is all used " ... " replace).
3, computing machine carries out association to first and second individual character phonetic " w ǒ m é n＜I door〉", and step is: (1) searches w ǒ dictionary in pinyin lexicon; (2) in w ǒ dictionary, search " w ǒ m é n ", find; (3) " w ǒ m é n " corresponding character " we " is confirmed.
4, computing machine to " w ǒ m é n z à i＜we〉" carry out association (in w ǒ dictionary, seek " w ǒ m é n z à i " under " w ǒ m é n ", with determine＜my door exists whether vocabulary or phrase), no result.
5, computing machine carries out association, no result to " z à i ch ī＜eating〉".(because＜my door〉confirm as " we " by association, computing machine can be not right again＜door exists〉carry out association.)
6, computing machine carries out association, no result to " ch ī ch á＜eat tea〉".
7, computing machine carries out association to " ch á di ǎ n＜refreshment〉", obtains two kinds of possible outcomes: " refreshment " and " making an inventory of " (two speech all belong to ch á dictionary " ch á dian " bar now).
8, computing machine directly shows the word of confirming through association, and the word of failing to confirm is kept the vocal print method
000 demonstration.So the result that screen display goes out is: ' our z à i (...) ch ī eat (...) refreshment 1 ↑ (making an inventory of) '.(part in ' ' is a screen display content, down together.Confirmed speech and phrase to call the association of this process in the following text, but confirmed finally that not the sentence of full content is a sentence to be confirmed.)
9, the user confirms (method such as vocal print method) to unacknowledged word in the sentence to be confirmed: by 0 or acknowledgement key, " " word is identified, cursor " ↑ " moves under " 0 position " of next sequence to be confirmed (promptly " eating " under the word) automatically; By 0 or acknowledgement key, " eating " word is identified, and cursor moves under " tea " word; By 0 or acknowledgement key, " refreshment " is identified.So far, whole sentence is confirmed to finish, and screen display changes into: ' we are eating refreshment '.
↑
Illustrate:
1, behind the use phonetic association method, the user will want that the whole sentences and phrases of the literal imported go into, and allow computing machine confirm a part automatically, and computing machine is parked in cursor " ↑ " under " 0 position " of the sequence to be confirmed of first word sound unconfirmed then, confirm that Deng the user confirmation method is with the vocal print method.
2, the sequence to be confirmed at cursor place has been confirmed one of them phonetically similar word as the user, cursor moves to next sequence to be confirmed automatically, and can not point to confirmed word, confirms under the word unless the user moves to cursor.
3, will change confirmed word as the user, available conventional method is eliminated word, re-enters.If but the word sound is constant, just the phonetically similar word selection is wrong, also cursor can be moved under this word, and by ad hoc " resetkey ", computing machine is about to phonetic and a series of phonetically similar word of this word and shows (demonstration that promptly recovers the vocal print method), reselects for the user.
0 1
4, for homonym, as " refreshment (making an inventory of) " just now, it is selective to show literal and position symbol (0,1,2 etc.) on the computing machine, does not show phonetic.
5, when using vocal print method, phonetic association method, the keyboard input simultaneously effectively.For " 0～9 ", English alphabet and various symbol (comprising punctuation mark), should use the keyboard input, and for exempting from confusion, the spy makes following regulation: ten numerals in (1) " 0～9 " must treat that cursor does not refer to can import (otherwise the meaning of numeral is to select the phonetically similar word of sequence to be confirmed relevant position) when sequence to be confirmed, so generally should all confirm to finish input digit again in sequence to be confirmed; (2) for ",. : " these seven punctuation marks, also should be preceding and import when not having sequence to be confirmed, confirm principle (as follows and) otherwise must meet whole sentence.
6, read softly in " " word " dictionary ", but many people custom is read as m é n, from practicality, should allow two kinds of pronunciations, the word of other analogues also can be with reference to this method, but is preceding topic not cause confusion.
For further raising the efficiency, the phonetic association method should increase following some content:
1, whole sentence method of ascertainment: after the user is read in short, if " 0 position " that word unconfirmed in the sentence to be confirmed that is left after association and speech all are in sequence to be confirmed, then available ",. : " in these seven punctuation marks any one, whole sentence is confirmed.Above example just meet this situation, when the user will＜we are eating refreshment read the computing machine warp to computing machine
After 0001 associations result that provides be ' our z à i (...) ch ī eat (...) refreshment (making an inventory of) ', this
As long as it is strong that ↑ time clicks fullstop, the demonstration of screen changes promptly that ' we are eating refreshment into.’。Just
↑ say that key in above-mentioned seven punctuation marks any one, the sequence to be confirmed of all word sounds unconfirmed is confirmed all selected 0 position word on the screen.
2, unique word is confirmed: partial words waits, do not have other phonetically similar words, user to read in these word sound computer-chronographs as " twisting (n ǐ ng) ", " girl (ni ū) " should directly provide corresponding literal (this word has more than 200 individual in " dictionary ").
3, unique everyday character is confirmed: part phonetic bar phonetically similar word now only has one to be everyday character in " dictionary ", and all the other phonetically similar words are seldom used.Allowing computing machine directly unique everyday character be confirmed when hearing this part word sound, also is a kind of way of raising the efficiency.What want to import as the user is the phonetically similar word of few usefulness really, rather than that everyday character, can utilize resetkey to find the word of wanting.For example (w ǒ) bar have now " I,
" two words, use and the latter is obviously few.
Illustrate: though unique space, unique everyday character can confirm directly that their phonetic also should be participated in association, in order to avoid the result of influence association.
Give some instances again now, further specify associative process polyphonic word, phrase.At first will for example be " absolute value " three words, the user reads＜absolute value 〉, the vocal print method should provide " ju é exhausted (...) du ì to (...) zh í value (...) ".In fact, the phonetic association method just carries out association to phonetic, and ignores literal, and problem for convenience of description below is saved sequence to be confirmed when needed, and the vocal print method is called " treating association's pinyin-group " according to the pinyin string that user's pronunciation provides.
Example 2, the user reads＜absolute value 〉. the vocal print method provides treats association's pinyin-group " ju é du ì zh í ", associative process: (1) looks for " ju é du ì " in ju é dictionary, finds, and the absolute ＂ of literal ＂ behind the ＂ ju é du ì ＂ is confirmed to the ＂ ju é du ì ＂ in the pinyin-group; (2) in ju é dictionary, look for ＂ ju é du ì zh í ＂, find, confirm (behind the absolute ＂ of ＂, increasing ＂ value ＂ word), provide literal at last: absolute value.
Example 3, the user is read in＜absolute gentleness 〉, the association's pinyin-group for the treatment of that provides by the vocal print method is " ju é du ì w ē n h é ", associative process: earlier first and second single-tone phonetic is carried out association, " ju é du ì " arranged in the ju é dictionary, provide corresponding character " definitely "; Again first, second and third single-tone phonetic is carried out association, discovery is the part in (ju é du ì w ē n d ù) little clauses and subclauses, then conversely to treating to look for " ju é du ì w ē n d ù ", no result in association's pinyin-group; Then third and fourth single-tone phonetic " w ē n h é＜gentleness〉" is carried out association, the result is arranged, provide corresponding literal " gentleness ".The overall result that provides at last is " absolute gentle ".
Example 4, the user is read in＜Five Principles of Peaceful Coexistence 〉, the vocal print method provides " h é p í ng g ò ngch ǔ w ǔ xi à ng yu á n z é ", associative process: (1) associates first and second individual character phonetic, promptly in h é dictionary, look for (h é p í ng), have, it is confirmed into literal " peace "; (2) first, second and third individual character phonetic is carried out association, discovery is the part of (h é p í ng g ò ng ch ǔ) little clauses and subclauses, so conversely to treating to seek this pinyin string (referring to h é p í ng g ò ngch ǔ) in association's pinyin-group, have, confirm to become " peaceful coexistence " (because of having confirmed " peace " in the front, so only need to increase " coexistence " two words); (3) to first to five totally five individual character phonetics carry out association, discovery is the part of (h é p í ng g ò ng ch ǔ w ǔ xi à ng yu á n z é) little clauses and subclauses, then conversely to the pinyin string for the treatment of to look in association's pinyin-group these little clauses and subclauses, have, confirm into literal (behind former result, increasing " five principles ").The result that last computing machine association goes out is: the Five Principles of Peaceful Coexistence.
Generally speaking, the process of association is exactly to find out speech phonetic and the phrase phonetic for the treatment of in association's pinyin-group, thereby the speech in the statement that the user is read in is directly confirmed into literal with phrase and (then in order homonym is listed when homonym occurring, select by the user), so just given full play to the function of computing machine, alleviated people's work, made the Chinese character input become simple, quick.
In Chinese sentence, the consumption of speech and phrase is sizable,, adds the help that " unique word affirmation ", " unique everyday character affirmation " reach " whole sentence is confirmed " three kinds of methods generally all more than 50%, estimates to want word that the user goes to select below 20%.
Also should replenish a bit at last: the function of the cursor of carrying " ↑ " is chosen position, affirmation literal in the literary composition, recover vocal print method demonstration (being used for resetkey) etc., be not parked in the end of article when it is many, so when practical application is of the present invention, keyboard input cursor should with herein ↑ cursor works simultaneously, and bears different functions respectively.
The invention has the advantages that the accuracy that has improved speech recognition, and the advantage of Chinese is given full play to.In fact, the advantage of Chinese is " by word group speech, word adds phrase and forms a complete sentence, and the everyday character sound about 1300 has just been formed thousands upon thousands works and expressions for everyday use altogether ".The present invention has brought into play this advantage of Chinese just, has avoided this shortcoming of Chinese character " font complexity " simultaneously.In whole process with vocal print method input Chinese character, the people only needs to read word with mouth, and with hand keypad word selection, the thinking that need not beat one's brains fully, the memory that puts the mind to, all complicated work are all gone to have finished by computing machine.
Though other language as English, Japanese etc., can be imported with the vocal print method too, its principle is the same, and concrete effect but may fall far short.Such as English, if by the defeated people of letter, situation is very simple, but the speed of input may be fast not as the hand kbhit, if press the speech input, situation can be very complicated, because the vocabulary of English is imitated in ten thousand.So English uses the vocal print method not have advantage (but the handicapped people of adversary is with the obvious advantage).
Realize the present invention, can consider following two kinds of schemes:
Scheme one: make a kind of " vocal print card " specially, contain " standard vocal print maker ", " sound and grain input method software and phonetic association method software " in the card, and enough storage unit are arranged for running software.An interface of card connects microphone, and another interface connects computing machine.Like this, general computing machine can both utilize vocal print card input Chinese character now, need not to increase storage unit.
Scheme two: utilize existing sound card processed voice, replace the function of " standard vocal print maker ", and " sound and grain input method software and phonetic association method software " is produced in the CD.This scheme requires computing machine that bigger memory capacity is arranged, and is equipped with sound card.
Claims (5)
1, computer method given in a kind of phonetic entry Chinese character of using, it is characterized in that each user must carry out " input of the standard that prestores vocal print sign indicating number " earlier, machine to be calculated has been remembered after user's the vocal print feature of whole Chinese everyday characters pronunciation, the user can use the phonetic entry Chinese character, at this moment, the user is read the vocal print feature of word sound with computing machine and whole word sound samples that the he or she prestores compare, thereby determine the content of the herringbone sound of reading, and provide corresponding phonetic and serial phonetically similar word, select for the user.
2, a kind of method of utilizing association to accelerate Chinese character input speed, it is characterized in that to treat that phonetic and the speech phonetic in the pinyin lexicon, phrase phonetic in association's pinyin-group compare, thereby find out speech phonetic, the phrase phonetic treated in association's pinyin-group, and they are directly confirmed into corresponding character (when homonym occurring, then whole homonyms are listed and numbered, select by the user), simultaneously, also utilize whole sentence method of ascertainment, unique everyday character to confirm two kinds of confirmation methods, further accelerate the speed of Chinese character input.
3, as the pinyin lexicon in the method as described in the claim 2, it is characterized in that 1332 single-tone phonetics arrange in alphabetical order the big layout of formation pinyin lexicon, the speech phonetic that is all started by unisonance under the single-tone phonetic, phrase phonetic are as little clauses and subclauses, these little clauses and subclauses arrange in alphabetical order, and speech, phrase be immediately following becoming only word content in the little clauses and subclauses after their phonetic, and can there be following several respects in the source of speech and phrase:
(1) speech and phrase in " modern Chinese dictionary " (78 years versions);
(2) new term of recent two decades appearance, new predicate;
(3) vocabulary and the phrase of professional, professional;
(4) user needs the speech and the phrase of input at any time by oneself.
4, as the whole sentence method of ascertainment in the method as described in the claim 2, it is characterized in that available ",. : " in these seven punctuation marks any one, will meet whole sentence and confirm that the sentence to be confirmed of condition carries out whole sentence and confirms, and the condition that whole sentence is confirmed is: the word unconfirmed in the sentence to be confirmed is all in " 0 position " of sequence to be confirmed.
5, confirm as the unique everyday character in the method as described in the claim 2, it is characterized in that if only have one to be everyday character in the sequence to be confirmed of certain word sound, remaining word is few usefulness at ordinary times, and then computing machine is confirmed this everyday character automatically, and adopt the way of resetkey to select the word of all the other few usefulness.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 98114665 CN1210301A (en) | 1997-07-11 | 1998-06-30 | Phonetical vein inputting method and phonetical letter association method |
Applications Claiming Priority (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 97112848 CN1178940A (en) | 1997-07-11 | 1997-07-11 | Computer sound and grain input method |
| CN97112848.0 | 1997-07-11 | ||
| CN97125316 | 1997-12-16 | ||
| CN97125316.1 | 1997-12-16 | ||
| CN 98114665 CN1210301A (en) | 1997-07-11 | 1998-06-30 | Phonetical vein inputting method and phonetical letter association method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1210301A true CN1210301A (en) | 1999-03-10 |
Family
ID=27179128
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 98114665 Pending CN1210301A (en) | 1997-07-11 | 1998-06-30 | Phonetical vein inputting method and phonetical letter association method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1210301A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100419647C (en) * | 2004-03-29 | 2008-09-17 | 台达电子工业股份有限公司 | Chinese-character-unit speech-sound inputting method and system |
| CN1901041B (en) * | 2005-07-22 | 2011-08-31 | 康佳集团股份有限公司 | Voice dictionary forming method and voice identifying system and its method |
| CN101458928B (en) * | 2007-12-10 | 2011-11-02 | 富士通株式会社 | Voice recognition apparatus |
| CN108519870A (en) * | 2018-03-29 | 2018-09-11 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN108664984A (en) * | 2017-03-28 | 2018-10-16 | 深圳市凯立德科技股份有限公司 | A kind of method and device of data inspection |
| CN111597531A (en) * | 2020-04-07 | 2020-08-28 | 北京捷通华声科技股份有限公司 | Identity authentication method and device, electronic equipment and readable storage medium |
-
1998
- 1998-06-30 CN CN 98114665 patent/CN1210301A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100419647C (en) * | 2004-03-29 | 2008-09-17 | 台达电子工业股份有限公司 | Chinese-character-unit speech-sound inputting method and system |
| CN1901041B (en) * | 2005-07-22 | 2011-08-31 | 康佳集团股份有限公司 | Voice dictionary forming method and voice identifying system and its method |
| CN101458928B (en) * | 2007-12-10 | 2011-11-02 | 富士通株式会社 | Voice recognition apparatus |
| CN108664984A (en) * | 2017-03-28 | 2018-10-16 | 深圳市凯立德科技股份有限公司 | A kind of method and device of data inspection |
| CN108664984B (en) * | 2017-03-28 | 2024-04-09 | 深圳市凯立德科技股份有限公司 | Data checking method and device |
| CN108519870A (en) * | 2018-03-29 | 2018-09-11 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN111597531A (en) * | 2020-04-07 | 2020-08-28 | 北京捷通华声科技股份有限公司 | Identity authentication method and device, electronic equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR100656736B1 (en) | System and method for disambiguating phonetic input | |
| JP4829901B2 (en) | Method and apparatus for confirming manually entered indeterminate text input using speech input | |
| CN1260704C (en) | Method for voice synthesizing | |
| CN1918578B (en) | Handwriting and voice input with automatic correction | |
| US6876967B2 (en) | Speech complementing apparatus, method and recording medium | |
| CN1232226A (en) | Sentence processing apparatus and method thereof | |
| CN1315809A (en) | Apparatus and method for spelling speech recognition in mobile communication | |
| US4468756A (en) | Method and apparatus for processing languages | |
| CN1910573A (en) | System for identifying and classifying denomination entity | |
| JP2008517399A5 (en) | ||
| CN1942875A (en) | Dialogue supporting apparatus | |
| Lieberman et al. | How to wreck a nice beach you sing calm incense | |
| CN101067766A (en) | Method for cancelling character string in inputting method and word inputting system | |
| TWI295783B (en) | Text inputting device for mobile communication device and method thereof | |
| CN1210301A (en) | Phonetical vein inputting method and phonetical letter association method | |
| CN101577115A (en) | Voice input system and voice input method | |
| JP5701327B2 (en) | Speech recognition apparatus, speech recognition method, and program | |
| CN101046706A (en) | Universal input method for different person computer and mobile phone | |
| CN111429886B (en) | Voice recognition method and system | |
| CN1110738C (en) | Literal character input method for notobook computer | |
| JP2021193608A (en) | Utterance generation device, utterance generation method, and computer program | |
| CN1607492A (en) | Digital electronic device and bopomofo input method using the same | |
| CN1084500C (en) | Chinese characters alternating device | |
| CN1257445C (en) | Chinese-character 'Pronunciation-meaning code' input method | |
| Willis et al. | A probabilistic flexible abbreviation expansion system for users with motor disabilities |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |