CN101082836A - Chinese characters input system integrating voice input and hand-written input function - Google Patents
Chinese characters input system integrating voice input and hand-written input function Download PDFInfo
- Publication number
- CN101082836A CN101082836A CN 200710052608 CN200710052608A CN101082836A CN 101082836 A CN101082836 A CN 101082836A CN 200710052608 CN200710052608 CN 200710052608 CN 200710052608 A CN200710052608 A CN 200710052608A CN 101082836 A CN101082836 A CN 101082836A
- Authority
- CN
- China
- Prior art keywords
- chinese character
- hand
- written
- voice
- module
- 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
Images
Abstract
The invention discloses a Chinese character inputting system to integrate the voice input and hand-written input function in the applying technical domain of Chinese character inputting method, which comprises the following parts: hand-written/voice input mode, hand-written/voice signal character generating mode, hand-written/voice candidate Chinese character array generating mode, integrating mode, Chinese character identifying and displaying mode, hand-written/voice model exercising mode and hand-written/voice database. The invention improves the identifying rate of Chinese character during hand-written input with voice input as accessory, which accelerates the identifying step without changing the custom of user.
Description
Technical field
The invention belongs to the Computer Recognition Technology field, it is the Chinese character input system of a kind of integrating speech sound input and hand-write input function, this system can improve the discrimination of input Chinese character, helps the consumer to use input systems such as mobile phone, computer to carry out the Chinese character input more easily.
Background technology
Along with developing rapidly of mobile communication technology, the quantity of mobile phone, computer user is also in quick growth, and the input method that is applied on mobile phone, the computer is also more and more important.Because the number of keys on the mobile phone, computer is limited, input information is often wasted time and energy on limited button, needs usually constantly to search up and down in the screen menu just can obtain to want the information imported, makes that the manual typing of traditional literal is more loaded down with trivial details.And the button size on mobile phone, the computer is all smaller usually, and many elderlys make that owing to the reason of eyesight does not see the letter on Chu's button or can skillfully not use phonetic/stroke input method manually the input Chinese character is quite inconvenient.At above problem, there is the people just to propose pronunciation inputting method and the application of hand-written inputting method on input systems such as mobile phone and computer.
In the technical development of voice and handwriting input identification, its correlation technique has been seen in all kinds of technological documents, for example with speech recognition, publication number is the apparatus and method that proposed to have proposed in the patent documentation that a kind of speech recognition input method of Chinese character, publication number are CN1494299A a kind of converting speech sound input into characters on handset in the patent documentation of CN1373406A.On the other hand, handwriting recognition is then as providing a kind of mobile phone with input, demonstration and transfer function of hand-written character in the patent documentation of publication number for CN1335703A, be method and the corresponding system that has proposed the handwriting input of a kind of mobile phone character in the patent documentation of CN1350390A perhaps, it serves to show that the recognition technology of voice and handwriting input is ripe all gradually as publication number.
Yet, though above-mentioned each patented technology is all in the improvement of algorithm, the feature extraction of hand-written/phonetic entry, or a lot of work has been done in criterion or the like aspect of setting up of improving voice or hand-written model, the raising of right its discrimination is still limited, because some limitation of the recognition technology of handwriting input, make for the extremely similar Chinese character of structure, as " the sixth of the twelve Earthly Branches ", " " and " own " and " day " and " saying ", even " four " and " ware ", " river " and " red " etc. all can't be discerned or can't discern well; For phonetic entry, it is more obvious that this limitation shows, because the Chinese character of same sound correspondence often has a lot of, which is so on earth? said method can't address this problem equally.
In view of said method very limited aspect the discrimination raising, the someone proposes to integrate hand-written and voice input signal to improve the notion of discrimination.
Wherein, publication number is the recognition methods that the patent documentation " recognition methods of integrating speech sound and handwriting input and system " of CN1549244A has proposed a kind of integrating speech sound and handwriting input, this method is after reception one has the voice/handwriting input of a character earlier, discern this voice/handwriting input and produce word row, and these word row have a plurality of and the corresponding identification character of this character, receive a hand-written/phonetic entry of describing one of this character feature then, the last identification character that is consistent most by acquisition and this character in this character row according to this feature, therefore integrating speech sound and handwriting input effectively, and improve discrimination by this.
Yet,, still have many problems to be solved though said method really can reach its effect in the raising of discrimination.As:
What (1) said method proposed must distinguish first input and second input, i.e. first input be a phonetic entry and a hand-written input one of them, second input be this phonetic entry and handwriting input wherein another, the method of its identification character is, produce word row according to first input, receive second input then, and according to the identification character that captures in the word row of second input by the first input generation and second input is consistent most, the problem of Cun Zaiing is like this: if first when being input as handwriting input, just can't finishing first input when user runs into the word that can not write and produce word row; If first when being input as phonetic entry, also can't finishing first input when user runs into the word that can not read and produce word row; Like this, can't produce a word according to first input and be listed as, also just can't import in the word row that produce by first and capture the identification character that is consistent most with second input, thereby can't reach the purpose of correct input according to second input;
(2) only just work under training mode of the hand-written/speech model training aids of said method proposition, promptly before this system of use, specially it is trained in advance by the user, so just can reach hand-written/speech model of setting up the individual and be stored in purpose in hand-written/speech database, and in use, this hand-written/speech database remains unchanged, the training patterns of this static state is more single, does not also meet user's use habit simultaneously.
Summary of the invention
The object of the present invention is to provide the Chinese character input system of a kind of integrating speech sound input and hand-write input function, this system can help the consumer to use input systems such as mobile phone, computer to carry out the input of Chinese character more easily, has solved the problem that above-mentioned existing method exists simultaneously preferably.
The Chinese character input system of a kind of integrating speech sound input provided by the invention and hand-write input function comprises hand-written model training module, hand-written data storehouse, handwriting input module, hand-written signal feature generation module and handwritten form candidate Chinese character word column-generation module;
The handwriting input module is used to receive the handwriting input of Chinese character, sends hand-written model training module and hand-written signal feature generation module respectively to after being converted into digital signal;
Hand-written model training module is used for adjusting the parameter in hand-written data storehouse, and it is stored to the hand-written data storehouse;
The hand-written data storehouse is used to store handwritten Chinese character and the matched data of Chinese character and relevant matched rule, receives and handle the request that handwritten form candidate Chinese character word column-generation module sends;
Hand-written signal feature generation module is used for extracting effective hand-written signal feature from digital signal, and it is sent into handwritten form candidate Chinese character word column-generation module;
After handwritten form candidate Chinese character word column-generation module receives the hand-written signal feature of the handwritten form to be identified that hand-written signal feature generation module generates, send request to the hand-written data storehouse, require the hand-written data storehouse that the hand-written signal feature of its all Chinese characters of having stored is provided, after receiving the hand-written signal feature of its all Chinese characters of having stored that the hand-written data storehouse provides, calculate the similarity size between the hand-written signal feature of the hand-written signal feature of handwritten form to be identified and its all Chinese characters of having stored that the hand-written data storehouse provides, produce handwritten form candidate Chinese character word row;
It is characterized in that: this system also comprises voice input module, phonic signal character generation module, voice candidate Chinese character word column-generation module, speech database, speech model training module, integrate module and Chinese Character Recognition and display module;
Voice input module is used to receive the phonetic entry of maximum Chinese characters, and the output of this speech input device is converted into digital signal, sends phonic signal character generation module and speech model training module respectively to;
The phonic signal character generation module is used to extract effective phonic signal character, and is sent to voice candidate Chinese character word column-generation module;
After voice candidate Chinese character word column-generation module receives the phonic signal character of the voice to be identified that the phonic signal character generation module generates, send request to speech database, after receiving the phonic signal character of its feedback, calculate the similarity size between the phonic signal character of the phonic signal character of voice to be identified and its all Chinese characters of having stored that speech database provides, be used to produce voice candidate Chinese character word row, be sent to integrate module;
Speech database is used to store voice and the matched data of Chinese character and relevant matched rule etc., is used to receive the also request of processed voice candidate Chinese character word column-generation module transmission;
The speech model training module is used for adjusting the parameter of speech database, and is stored to speech database;
Handwritten form candidate Chinese character word row that integrate module reception handwritten form candidate Chinese character word column-generation module and voice candidate Chinese character word column-generation module are sent and voice candidate Chinese character word row, the two is made up, produce whole candidate Chinese character word row, send it to Chinese Character Recognition and display module;
Chinese Character Recognition and display module are used for selecting from the whole candidate Chinese character word row that integrate module produces the Chinese character of weight maximum, and this Chinese character is shown that if sky classified as in whole candidate Chinese character word, then prompting is re-entered.
When system of the present invention uses, be aided with phonetic entry when receiving the handwriting input of a Chinese character, when having improved the Chinese Character Recognition rate, solved above-mentioned owing to show the highest Chinese character of a frequency according to first input signal earlier, import again that secondary signal is used to revise and the trouble brought, when also having solved handwriting input as first input Chinese character can not write or phonetic entry during as first input phonetic piece together inaccurate and can't finish the problem of input and independent when only using handwriting input or phonetic entry, the problem that can't discern or can't correctly discern structural similarity or the identical Chinese character that pronounces; Simultaneously, utilize hand-written/speech model training module, use in the process of this input system the individual, set up hand-written/speech model of individual, and be stored in hand-written/speech database, quickened identification step carrying out, improved discrimination, and it dynamically updates hand-written/speech database in the process of using, needn't artificially train it specially in advance, meet people's use habit, especially fast effectively for the identification of the Chinese character of high-repetition-rate.
Description of drawings
Fig. 1 is the structural representation of system of the present invention;
Fig. 2 is for Chinese Character Recognition of the present invention and display module is listed as according to handwritten form candidate Chinese character word and the assembled state of voice candidate Chinese character word row is carried out the assembled state figure that corresponding Chinese character shows or prompting is re-entered;
Fig. 3 is the workflow diagram of the embodiment of the invention.
Embodiment
As shown in Figure 1, the Chinese character input system of a kind of integrating speech sound input of the present invention and hand-write input function comprises hand-written model training module 1, hand-written data storehouse 2, handwriting input module 3, hand-written signal feature generation module 4, handwritten form candidate Chinese character word column-generation module 5, integrate module 6, Chinese Character Recognition and display module 7, voice input module 8, phonic signal character generation module 9, voice candidate Chinese character word column-generation module 10, speech database 11, speech model training module 12.
Hand-written model training module 1 is used for using the individual process of this Chinese character input system, constantly adjust the parameter in the hand-written data storehouse, make the performance of hand-written model constantly approach to certain optimum condition (as: all Chinese characters in the hand-written data storehouse are had best discrimination), thereby set up individual's hand-written model, the personal handwritten model can wait by movable contour model, based on the deformable elastic matching template of stroke and SVM (support vector) and set up.
Hand-written data storehouse 2 stores handwritten Chinese character and (as: hand-written signal features etc.) such as the matched data of Chinese character and relevant matched rules, be used to receive and handle the request that handwritten form candidate Chinese character word column-generation module 5 sends, and the parameter in the hand-written data storehouse is dynamically adjusted in the process of hand-written model training, makes the performance of system model constantly approach to certain optimum condition (as: all Chinese characters in the hand-written data storehouse are had best discrimination);
Handwriting input module 3 is used to receive the handwriting input of a Chinese character, comprises a hand input device (as: handwriting pad), writes down the position of corresponding center of effort by this hand input device, and is translated into digital signal;
Hand-written signal feature generation module 4 is used for extracting effective hand-written signal feature from digital signal, and it is sent into handwritten form candidate Chinese character word column-generation module 5, here the effective hand-written signal feature of Ti Quing is in the feature that can distinguish this Chinese character and other Chinese characters (architectural feature or statistical nature) through obtaining on the pretreated basis, as: the stroke feature that can be based on whole word, it also can be the radical feature, it can also be person's handwriting distribution statistics feature of Chinese character etc., according to particular problem, pre-service can comprise binaryzation, smoothing denoising, slant correction, curve fitting, the pen section is optimized, delete invalid pen section, the normalization of pen section, steps such as non-linear normalizing;
After handwritten form candidate Chinese character word column-generation module 5 receives the hand-written signal feature of the handwritten form to be identified that hand-written signal feature generation module 4 generates, send request to hand-written data storehouse 2, require hand-written data storehouse 2 that the hand-written signal feature of its all Chinese characters of having stored is provided, after receiving the hand-written signal feature of its all Chinese characters of having stored that hand-written data storehouse 2 provides, calculate the similarity size between the hand-written signal feature of the hand-written signal feature of handwritten form to be identified and its all Chinese characters of having stored that hand-written data storehouse 2 provides, be used to produce handwritten form candidate Chinese character word row (promptly realizing the Chinese character coupling), these handwritten form candidate Chinese character word row have a plurality of and the corresponding identification Chinese character of this handwritten form to be identified, each identification Chinese character is composed with weight according to its similarity size, this word is listed as according to the weight of identification Chinese character and arranges from big to small, difference (architectural feature or statistical nature) according to the hand-written signal feature of extracting, the method that realizes the Chinese character coupling is also different, wherein, Chinese character matching process based on architectural feature has: based on the pen section, the template of stroke is put in order the word matching method, lax coupling, structured analysis method,, attributed relational graph (ARG) method, deformable elasticity stroke Matching Model or the like; Chinese character matching process based on statistical nature has: neural network, FCM rough sort and SVM (support vector machine) etc.;
Integrate module 6 produces whole candidate Chinese character word row according to the assembled state of handwritten form candidate Chinese character word row that produce and voice candidate Chinese character word row, integration method is: when handwritten form candidate Chinese character word row are sky simultaneously with voice candidate Chinese character word row, sky classified as in whole candidate Chinese character word, gives Chinese Character Recognition and display module 7 with whole candidate Chinese character word biographies; When handwritten form candidate Chinese character word row non-NULL, when sky classified as in voice candidate Chinese character word, handwritten form candidate Chinese character word is listed as a whole candidate Chinese character word biographies gives Chinese Character Recognition and display module 7; When sky classified as in handwritten form candidate Chinese character word, during voice candidate Chinese character word row non-NULL, voice candidate Chinese character word is listed as a whole candidate Chinese character word biographies gives Chinese Character Recognition and display module 7; When handwritten form candidate Chinese character word row and voice candidate Chinese character word row while non-NULL, and when identical Chinese character is arranged in two candidate row, the Chinese character that all are identical is arranged from big to small by the handwritten form weight, the Chinese Character after the ordering is listed as a whole candidate Chinese character word biographies gives Chinese Character Recognition and display module 7; When handwritten form candidate Chinese character word row and voice candidate Chinese character word row while non-NULL, and when not having identical Chinese character in two candidate row, according to the handwritten form principle of priority handwritten form candidate Chinese character word is listed as a whole candidate Chinese character word biographies and gives Chinese Character Recognition and display module 7, wherein, hand-written/phonetic entry shared weight in Chinese Character Recognition is meant that the handwritten form/voice of each candidate Chinese character and this user input relative to each other are the possibility degree (probability, similarity size) that meets;
Chinese Character Recognition and display module 7 are used for selecting from the whole candidate Chinese character word row that integrate module 6 produces the Chinese character of weight maximum, and this Chinese character is shown that if sky classified as in whole candidate Chinese character word, then prompting is re-entered;
Voice input module 8 is used to receive the phonetic entry of a no more than Chinese character, comprises a speech input device (as: microphone), and the output of this speech input device is converted into digital signal;
Phonic signal character generation module 9 is used for removing voice to discerning inessential redundant information, extract effective phonic signal character, here the effective phonic signal character of Ti Quing is at the syllable characteristic that passes through the Chinese character that obtains on the pretreated basis, it also can be tone feature of Chinese character etc., phonic signal character commonly used has MFCC (Mel frequency cepstral coefficient), LPCC (linear prediction cepstrum coefficient) or the like, according to particular problem, pre-service can comprise steps such as branch frame, end-point detection, pre-emphasis, segmentation, windowing;
After voice candidate Chinese character word column-generation module 10 receives the phonic signal character of the voice to be identified that phonic signal character generation module 9 generates, send request to speech database 11, require speech database 11 that the phonic signal character of its all Chinese characters of having stored is provided, after receiving the phonic signal character of its all Chinese characters of having stored that speech database 11 provides, calculate the similarity size between the phonic signal character of the phonic signal character of voice to be identified and its all Chinese characters of having stored that speech database 11 provides, be used to produce voice candidate Chinese character word row (promptly realizing speech recognition), these voice candidate Chinese character word row have a plurality of and the corresponding identification Chinese character of these voice to be identified, each identification Chinese character is composed with weight according to its similarity size, this word is listed as according to the weight of identification Chinese character and arranges from big to small, is used to realize that the method for speech recognition mainly contains dynamic time consolidation technology (DTW), vector quantization technology (VQ), hidden Markov model (HMM), artificial neural network (ANN) etc.;
Speech database 11 stores voice and the matched data of Chinese character and relevant matched rule etc. (as: frequency of utilization of all Chinese characters that phonic signal character and same sound are corresponding etc.), be used to receive the also request of processed voice candidate Chinese character word column-generation module 10 transmissions, and the parameter in the speech database is dynamically adjusted in the process of speech model training, makes the performance of system model constantly approach to certain optimum condition (as: voice to all Chinese characters in the speech database have best discrimination);
Speech model training module 12 is used for using the individual process of this Chinese character input system, constantly adjust the parameter in the speech database 11, make the performance of speech model constantly approach to certain optimum condition (as: voice to all Chinese characters in the speech database 11 have best discrimination), thereby set up individual's speech model, individual's speech model can be set up in several ways, as: dynamic time consolidation technology (DTW), hidden Markov model (HMM) and artificial neural network (ANN) or the like.
When system of the present invention uses, receive the handwriting input of a Chinese character and the phonetic entry (phonetic entry can be sky) of a no more than Chinese character by hand-written/voice input module, hand-written/pronunciation extracting module then carries out pre-service to above-mentioned input respectively, and therefrom extract corresponding validity feature, hand-written/after voice candidate Chinese character word column-generation module receives the validity feature of hand-written/handwritten form/voice to be identified that pronunciation extracting module extracted, send request to hand-written/speech database, require hand-written/speech database that the hand-written/phonic signal character of its all Chinese characters of having stored is provided, after the request of hand-written/hand-written/voice candidate Chinese character word column-generation module of speech database reception it is handled, hand-written/the phonic signal character of its all Chinese characters of having stored that hand-written/speech database is provided sends hand-written/voice candidate Chinese character word column-generation module to, hand-written/voice candidate Chinese character word column-generation module is calculated the similarity size between the hand-written/phonic signal character of the phonic signal character of hand-written signal feature/voice to be identified of handwritten form to be identified and its all Chinese characters of having stored that hand-written/speech database provides, the corresponding one handwritten form/voice candidate Chinese character word row that produce, and two word biographies are delivered to integrate module, integrate module then produces whole candidate Chinese character word row that sort from big to small according to weight according to the assembled state of two candidate Chinese character word row, at last, Chinese Character Recognition and display module extract the Chinese character of weight maximum in the whole candidate Chinese character word row and with its demonstration.
Because Chinese character has following characteristic: the pronunciation of the Chinese character that font (radicals by which characters are arranged in traditional Chinese dictionaries and stroke) is similar or identical is often different, and the font of the Chinese character that pronunciation is identical often difference is bigger.Therefore, integrate the Chinese character input method of handwriting input and phonetic entry, make the input that is aided with voice in input in the handwritten Chinese character, promptly can improve the discrimination of Chinese character effectively in conjunction with the characteristic of both complementations, also solved when first input limits, when the user can't finish the problem of input and only use handwriting input or phonetic entry individually because word can not be write or phonetic is pieced together inaccurate, the problem that can't discern or can't correctly discern structural similarity or the identical Chinese character that pronounces, simultaneously, the input mode of these two kinds of natures is compared to encode and is waited other input modes, hommization more, convenient.
Embodiment:
Example with the existing desire input of a user one " four " word comes enforcement of the present invention is described in further detail below, among this embodiment, adopt the recognition method of stroke-pen section-radical-whole word for the Chinese character of handwriting input, architectural feature with Chinese character---radical feature is as the hand-written signal feature, with the syllable characteristic of Chinese character as phonic signal character.
As shown in Figure 3, the course of work of this Chinese character input system is:
(1) user by handwriting pad import Chinese character "
", (in the time limit that allows) simultaneously, by microphone input voice " shi ", and will "
" send into hand-written model training module, " shi " sent into the speech model training module;
(2) hand-written signal feature generation module and phonic signal character generation module extract effective hand-written signal feature and phonic signal character, and its concrete steps comprise:
(2.1) concrete steps of hand-written signal feature extraction are as follows:
Steps such as (2.1.1) hand-written pre-service obtains a segment information of Chinese character through pre-service, and pre-service comprises that person's handwriting is level and smooth, person's handwriting normalization, stroke approach:
(2.1.1.1) person's handwriting is level and smooth: remove some isolated points, noise spot, get rid of disturbing factor, and fill by the point that difference will lack;
(2.1.1.2) person's handwriting normalization: handwriting is mapped in the middle of the fixing square frame, it is carried out yardstick, the operation of aspects such as angle;
(2.1.1.3) stroke is approached: go the matched curve section with broken line;
(2.1.2) radical extracts, and adopts FCM fuzzy clustering algorithm that the pen section is carried out automatic cluster and adopted additive method to reach the purpose of accurate extraction radical;
(2.1.3) hand-written signal feature extraction (being the radical Feature Extraction) is by adopting the mode of statistics to obtain the feature of radical to the radical that extracts;
(2.2) concrete steps of phonic signal character extraction are as follows:
(2.2.1) voice pre-service: divide frame, end-point detection, frame length can be chosen 20ms, and the end-point detection algorithm adopts the double threshold method based on short-time energy and zero-crossing rate;
(2.2.2) extract phonic signal character, feature is selected 12 dimension MFCC (Mel frequency cepstral coefficient) for use;
(3) with lax coupling with step (2.1) extract "
" the radical feature and the radical feature of each Chinese character in the database mate one by one; calculate the similarity between them; thus produce handwritten form candidate Chinese character word row that sort from big to small by weight: ware 60%, 4 30%, with 10% (, being " ware " (incorrect) then) according to handwriting recognition weight recognition result if having only handwriting input; With hidden Markov model (HMM) syllable characteristic of " shi " of step (2.2) extraction and the syllable characteristic of each Chinese character in the speech database are mated one by one, calculate the similarity between them, thereby produce voice candidate Chinese character word row that sort from big to small by weight: be 40%, make 30%, 10 12%, 4 10%, dead 8% (because this user pronunciation is inaccurate, if have only phonetic entry, be "Yes" (incorrect) then according to speech recognition weight recognition result), two word biographies are delivered to integrate module;
(4) integrate: the handwritten form candidate Chinese character word row " ware 60%, 4 30%, together 10% " that draw from step (3) are listed as " be 40%, make 30%, 10 12%, 4 10%, dead 8% " non-NULL simultaneously with voice candidate Chinese character word, and in two candidate row identical Chinese character " four " is arranged, according to Fig. 2, then that all are identical Chinese character is arranged from big to small by the handwritten form weight, with the Chinese Character row " 4 100% " after the ordering as a whole candidate Chinese character word biographies deliver to Chinese Character Recognition and display module;
(5) Chinese character " four " of weight selection maximum from the whole candidate Chinese character word row " 4 100% " that step (4) draws, and (promptly with its demonstration, if in the handwritten form input, be aided with phonetic entry " shi " (this sound is inaccurate), then integrating recognition result according to both is its identical Chinese character " four ", recognition result is consistent with the Chinese character of user expectation input, and identification is correct);
(6) training: adopt and to set up hand-written training pattern, and Chinese character root is trained based on the deformable elastic matching template of stroke, according to the handwritten form of step (1) input "
" and Chinese character " four " the change hand-written data storehouse identified of voice " shi " and step (5) in the radical feature of Chinese character " four ", make its with "
" the radical feature more close, then make the user import once more "
" time, "
" the radical feature and database in the similarity of radical feature of " four " bigger (as: can change as follows: ware 50%, 4 40%, with 10%); simultaneously; adopt hidden Markov model (HMM) to set up the voice training model; and Chinese character syllable trained, the syllable characteristic of Chinese character " four " in the change speech database (as: make behind the change syllable characteristic that the candidate Chinese character word row that calculate by similarity are as follows: 4 45%, dead 30%, thread 12%, be 10%, make 3%).
Claims (2)
1, the Chinese character input system of a kind of integrating speech sound input and hand-write input function comprises hand-written model training module (1), hand-written data storehouse (2), handwriting input module (3), hand-written signal feature generation module (4) and handwritten form candidate Chinese character word column-generation module (5);
Handwriting input module (3) is used to receive the handwriting input of Chinese character, sends hand-written model training module (1) and hand-written signal feature generation module (4) respectively to after being converted into digital signal;
Hand-written model training module (1) is used for adjusting the parameter of hand-written data storehouse (2), and it is stored to hand-written data storehouse (2);
Hand-written data storehouse (2) is used to store handwritten Chinese character and the matched data of Chinese character and relevant matched rule, receives and handle the request that handwritten form candidate Chinese character word column-generation module (5) sends;
Hand-written signal feature generation module (4) is used for extracting effective hand-written signal feature from digital signal, and it is sent into handwritten form candidate Chinese character word column-generation module (5);
After handwritten form candidate Chinese character word column-generation module (5) receives the hand-written signal feature of the handwritten form to be identified that hand-written signal feature generation module (4) generates, send request to hand-written data storehouse (2), require hand-written data storehouse (2) that the hand-written signal feature of its all Chinese characters of having stored is provided, after receiving the hand-written signal feature of its all Chinese characters of having stored that hand-written data storehouse (2) provides, calculate the similarity size between the hand-written signal feature of the hand-written signal feature of handwritten form to be identified and its all Chinese characters of having stored that hand-written data storehouse (2) provide, produce handwritten form candidate Chinese character word row;
It is characterized in that: this system also comprises voice input module (8), phonic signal character generation module (9), voice candidate Chinese character word column-generation module (10), speech database (11), speech model training module (12), integrate module (6) and Chinese Character Recognition and display module (7);
Voice input module (8) is used to receive the phonetic entry of maximum Chinese characters, and the output of this speech input device is converted into digital signal, sends phonic signal character generation module (9) and speech model training module (12) respectively to;
Phonic signal character generation module (9) is used to extract effective phonic signal character, and is sent to voice candidate Chinese character word column-generation module (10);
After voice candidate Chinese character word column-generation module (10) receives the phonic signal character of the voice to be identified that phonic signal character generation module (9) generates, send request to speech database (11), after receiving the phonic signal character of its feedback, calculate the similarity size between the phonic signal character of the phonic signal character of voice to be identified and its all Chinese characters of having stored that speech database (11) provides, be used to produce voice candidate Chinese character word row, be sent to integrate module (6);
Speech database (11) is used to store voice and the matched data of Chinese character and relevant matched rule etc., is used for receiving the also request of processed voice candidate Chinese character word column-generation module (10) transmission;
Speech model training module (12) is used for adjusting the parameter of speech database (11), and is stored to speech database (11);
Handwritten form candidate Chinese character word row that integrate module (6) reception handwritten form candidate Chinese character word column-generation module (5) and voice candidate Chinese character word column-generation module (10) are sent and voice candidate Chinese character word row, the two is made up, produce whole candidate Chinese character word row, send it to Chinese Character Recognition and display module (7);
Chinese Character Recognition and display module (7) are used for selecting from the whole candidate Chinese character word row that integrate module (6) produces the Chinese character of weight maximum, and this Chinese character is shown that if sky classified as in whole candidate Chinese character word, then prompting is re-entered.
2, Chinese character input system according to claim 1 is characterized in that: integrate module (6) makes up according to following method:
When handwritten form candidate Chinese character word row were sky simultaneously with voice candidate Chinese character word row, sky classified as in whole candidate Chinese character word, gives Chinese Character Recognition and display module (7) with whole candidate Chinese character word biographies;
When handwritten form candidate Chinese character word row non-NULL, when sky classified as in voice candidate Chinese character word, handwritten form candidate Chinese character word is listed as a whole candidate Chinese character word biographies gives Chinese Character Recognition and display module (7);
When sky classified as in handwritten form candidate Chinese character word, during voice candidate Chinese character word row non-NULL, voice candidate Chinese character word is listed as a whole candidate Chinese character word biographies gives Chinese Character Recognition and display module (7);
When handwritten form candidate Chinese character word row and voice candidate Chinese character word row while non-NULL, and when identical Chinese character is arranged in two candidate row, the Chinese character that all are identical is arranged from big to small by the handwritten form weight, the Chinese Character after the ordering is listed as a whole candidate Chinese character word biographies gives Chinese Character Recognition and display module (7);
When handwritten form candidate Chinese character word row and voice candidate Chinese character word row while non-NULL, and when not having identical Chinese character in two candidate row, according to the handwritten form principle of priority handwritten form candidate Chinese character word is listed as a whole candidate Chinese character word biographies and gives Chinese Character Recognition and display module (7).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710052608 CN101082836A (en) | 2007-06-29 | 2007-06-29 | Chinese characters input system integrating voice input and hand-written input function |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710052608 CN101082836A (en) | 2007-06-29 | 2007-06-29 | Chinese characters input system integrating voice input and hand-written input function |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101082836A true CN101082836A (en) | 2007-12-05 |
Family
ID=38912428
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200710052608 Pending CN101082836A (en) | 2007-06-29 | 2007-06-29 | Chinese characters input system integrating voice input and hand-written input function |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101082836A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102378951A (en) * | 2009-03-30 | 2012-03-14 | 符号技术有限公司 | Combined speech and touch input for observation symbol mappings |
| CN102708862A (en) * | 2012-04-27 | 2012-10-03 | 苏州思必驰信息科技有限公司 | Touch-assisted real-time speech recognition system and real-time speech/action synchronous decoding method thereof |
| CN101789073B (en) * | 2009-01-22 | 2013-06-26 | 富士通株式会社 | Character recognition device and character recognition method thereof |
| CN103218199A (en) * | 2013-02-26 | 2013-07-24 | 马骏 | Phonetic input method with identification code input function |
| CN103294370A (en) * | 2012-03-05 | 2013-09-11 | 北京千橡网景科技发展有限公司 | Method and equipment for triggering keystroke operation |
| CN104007925A (en) * | 2013-02-22 | 2014-08-27 | 三星电子株式会社 | Method and apparatus for making contents through writing input on touch screen |
| CN104166462A (en) * | 2013-05-17 | 2014-11-26 | 北京搜狗科技发展有限公司 | Input method and system for characters |
| CN104751142A (en) * | 2015-04-01 | 2015-07-01 | 电子科技大学 | Natural scene text detection algorithm based on stroke features |
| CN105590624A (en) * | 2014-11-10 | 2016-05-18 | 现代自动车株式会社 | Voice recognition system and method in vehicle |
| WO2016127619A1 (en) * | 2015-02-12 | 2016-08-18 | 中兴通讯股份有限公司 | Mixed input method and apparatus, and computer storage medium |
| CN105938558A (en) * | 2015-03-06 | 2016-09-14 | 松下知识产权经营株式会社 | Learning method |
| CN107391015A (en) * | 2017-07-19 | 2017-11-24 | 广州视源电子科技股份有限公司 | A kind of control method of Intelligent flat, device, equipment and storage medium |
| CN108197625A (en) * | 2017-12-18 | 2018-06-22 | 北京云星宇交通科技股份有限公司 | A kind of method and system for correcting Car license recognition |
| CN109643547A (en) * | 2016-08-31 | 2019-04-16 | 索尼公司 | Information processing unit, the methods and procedures for handling information |
| CN109791767A (en) * | 2016-09-30 | 2019-05-21 | 罗伯特·博世有限公司 | System and method for speech recognition |
| CN110049270A (en) * | 2019-03-12 | 2019-07-23 | 平安科技（深圳）有限公司 | Multi-person conference speech transcription method, apparatus, system, equipment and storage medium |
| CN110111781A (en) * | 2018-01-31 | 2019-08-09 | 丰田自动车株式会社 | Information processing unit and information processing method |
| CN111356977A (en) * | 2017-12-04 | 2020-06-30 | 深圳市柔宇科技有限公司 | Method for processing writing strokes and related equipment |
| CN112347288A (en) * | 2020-11-10 | 2021-02-09 | 北京北大方正电子有限公司 | Character and picture vectorization method |
| CN113687724A (en) * | 2021-07-23 | 2021-11-23 | 维沃移动通信有限公司 | Candidate character display method and device and electronic equipment |
-
2007
- 2007-06-29 CN CN 200710052608 patent/CN101082836A/en active Pending
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789073B (en) * | 2009-01-22 | 2013-06-26 | 富士通株式会社 | Character recognition device and character recognition method thereof |
| US9519353B2 (en) | 2009-03-30 | 2016-12-13 | Symbol Technologies, Llc | Combined speech and touch input for observation symbol mappings |
| CN102378951A (en) * | 2009-03-30 | 2012-03-14 | 符号技术有限公司 | Combined speech and touch input for observation symbol mappings |
| CN103294370A (en) * | 2012-03-05 | 2013-09-11 | 北京千橡网景科技发展有限公司 | Method and equipment for triggering keystroke operation |
| CN102708862A (en) * | 2012-04-27 | 2012-10-03 | 苏州思必驰信息科技有限公司 | Touch-assisted real-time speech recognition system and real-time speech/action synchronous decoding method thereof |
| CN102708862B (en) * | 2012-04-27 | 2014-09-24 | 苏州思必驰信息科技有限公司 | Touch-assisted real-time speech recognition system and real-time speech/action synchronous decoding method thereof |
| CN104007925B (en) * | 2013-02-22 | 2021-01-01 | 三星电子株式会社 | Method and apparatus for generating content through writing input on touch screen |
| CN104007925A (en) * | 2013-02-22 | 2014-08-27 | 三星电子株式会社 | Method and apparatus for making contents through writing input on touch screen |
| CN103218199A (en) * | 2013-02-26 | 2013-07-24 | 马骏 | Phonetic input method with identification code input function |
| CN104166462A (en) * | 2013-05-17 | 2014-11-26 | 北京搜狗科技发展有限公司 | Input method and system for characters |
| CN104166462B (en) * | 2013-05-17 | 2017-07-21 | 北京搜狗科技发展有限公司 | The input method and system of a kind of word |
| CN105590624A (en) * | 2014-11-10 | 2016-05-18 | 现代自动车株式会社 | Voice recognition system and method in vehicle |
| CN105590624B (en) * | 2014-11-10 | 2020-11-03 | 现代自动车株式会社 | Speech recognition system in vehicle and method thereof |
| WO2016127619A1 (en) * | 2015-02-12 | 2016-08-18 | 中兴通讯股份有限公司 | Mixed input method and apparatus, and computer storage medium |
| CN105938558A (en) * | 2015-03-06 | 2016-09-14 | 松下知识产权经营株式会社 | Learning method |
| CN105938558B (en) * | 2015-03-06 | 2021-02-09 | 松下知识产权经营株式会社 | Learning method |
| CN104751142B (en) * | 2015-04-01 | 2018-04-27 | 电子科技大学 | A kind of natural scene Method for text detection based on stroke feature |
| CN104751142A (en) * | 2015-04-01 | 2015-07-01 | 电子科技大学 | Natural scene text detection algorithm based on stroke features |
| CN109643547A (en) * | 2016-08-31 | 2019-04-16 | 索尼公司 | Information processing unit, the methods and procedures for handling information |
| CN109791767B (en) * | 2016-09-30 | 2023-09-05 | 罗伯特·博世有限公司 | System and method for speech recognition |
| CN109791767A (en) * | 2016-09-30 | 2019-05-21 | 罗伯特·博世有限公司 | System and method for speech recognition |
| CN107391015A (en) * | 2017-07-19 | 2017-11-24 | 广州视源电子科技股份有限公司 | A kind of control method of Intelligent flat, device, equipment and storage medium |
| CN107391015B (en) * | 2017-07-19 | 2021-03-16 | 广州视源电子科技股份有限公司 | Control method, device and equipment of intelligent tablet and storage medium |
| CN111356977A (en) * | 2017-12-04 | 2020-06-30 | 深圳市柔宇科技有限公司 | Method for processing writing strokes and related equipment |
| CN108197625A (en) * | 2017-12-18 | 2018-06-22 | 北京云星宇交通科技股份有限公司 | A kind of method and system for correcting Car license recognition |
| CN110111781A (en) * | 2018-01-31 | 2019-08-09 | 丰田自动车株式会社 | Information processing unit and information processing method |
| CN110111781B (en) * | 2018-01-31 | 2023-02-17 | 丰田自动车株式会社 | Information processing apparatus, information processing method, and computer program |
| CN110049270A (en) * | 2019-03-12 | 2019-07-23 | 平安科技（深圳）有限公司 | Multi-person conference speech transcription method, apparatus, system, equipment and storage medium |
| CN112347288A (en) * | 2020-11-10 | 2021-02-09 | 北京北大方正电子有限公司 | Character and picture vectorization method |
| CN112347288B (en) * | 2020-11-10 | 2024-02-20 | 北京北大方正电子有限公司 | Vectorization method of word graph |
| CN113687724A (en) * | 2021-07-23 | 2021-11-23 | 维沃移动通信有限公司 | Candidate character display method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101082836A (en) | Chinese characters input system integrating voice input and hand-written input function | |
| CN100349206C (en) | Text-to-speech interchanging device | |
| CN110136727B (en) | Speaker identification method, device and storage medium based on speaking content | |
| CN108346427A (en) | A kind of audio recognition method, device, equipment and storage medium | |
| CN102509547B (en) | Method and system for voiceprint recognition based on vector quantization based | |
| CN103000176B (en) | Speech recognition method and system | |
| CN101154380B (en) | Method and device for registration and validation of speaker's authentication | |
| CN107301865A (en) | A kind of method and apparatus for being used in phonetic entry determine interaction text | |
| CN109036391A (en) | Audio recognition method, apparatus and system | |
| CN108256458B (en) | Bidirectional real-time translation system and method for deaf natural sign language | |
| WO2020107834A1 (en) | Verification content generation method for lip-language recognition, and related apparatus | |
| CN108735200B (en) | Automatic speaker labeling method | |
| CN106796785A (en) | Sample sound for producing sound detection model is verified | |
| CN107093422B (en) | Voice recognition method and voice recognition system | |
| CN101540170B (en) | Voiceprint recognition method based on biomimetic pattern recognition | |
| CN1547191A (en) | Semantic and sound groove information combined speaking person identity system | |
| CN109378006A (en) | A kind of striding equipment method for recognizing sound-groove and system | |
| CN104808794A (en) | Method and system for inputting lip language | |
| WO2020238045A1 (en) | Intelligent speech recognition method and apparatus, and computer-readable storage medium | |
| CN112309365A (en) | Training method and device of speech synthesis model, storage medium and electronic equipment | |
| CN104427109A (en) | Method for establishing contact item by voices and electronic equipment | |
| CN101876887A (en) | Voice input method and device | |
| CN111445898A (en) | Language identification method and device, electronic equipment and storage medium | |
| CN101505328A (en) | Network data retrieval method applying speech recognition and system thereof | |
| CN101377726A (en) | Input method combining speech recognition with stroke recognition and terminal thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |