CN101266790A - Device and method for automatic time marking of text file - Google Patents
Device and method for automatic time marking of text file Download PDFInfo
- Publication number
- CN101266790A CN101266790A CNA2007100886277A CN200710088627A CN101266790A CN 101266790 A CN101266790 A CN 101266790A CN A2007100886277 A CNA2007100886277 A CN A2007100886277A CN 200710088627 A CN200710088627 A CN 200710088627A CN 101266790 A CN101266790 A CN 101266790A
- Authority
- CN
- China
- Prior art keywords
- time
- text file
- sentence
- voice
- algorithm
- 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 relates to a device and a method of word file automatically marking time. The method comprises: a receiving module receives the word file and an audio file, wherein, the word file is consisted of a plurality of sentences; a voice recognition module converts the sentences in the word file into a voice model, and divides the audio file into a plurality of frames according to interval time and numbers the frames in order, then changes speech data of the frames into characteristic parameter via capturing voice, and calculates a best voice route of the frame and the voice model which are matched with each other; a marking module captures a number of the frame corresponding to beginning of each sentence according to the best voice route, and the number of the frame and interval time gain starting time of the audio file corresponding to the beginning of each sentence, and mark the starting time on the work file. The method of the invention can make each sentence in the word file automatically mark the starting time corresponding to the audio file, and need not make use of artificial means to mark time sentence by sentence as traditional technology, so as to save a plurality of time and human.
Description
Technical field
The present invention relates to the apparatus and method that a kind of text file indicates the time, relate in particular to and a kind ofly carry out the apparatus and method that text file indicates the time automatically by speech recognition.
Background technology
No matter be language learner or speech player (for example, MP3 player), present most equipment all has the synchronous function of ci and qu.Just listening to that language is read aloud or during playback of songs as the user, having corresponding literal (in reading aloud perhaps the lyrics), following voice document and together playing.So that allow the user can listen to voice document on one side, Yi Bian read and voice document literal in correspondence with each other.So, when user's utilization has the learning equipment language of ci and qu synchronizing function or listens to song, can increase the efficient of language learning or quicken the efficient that song is learnt.
At present the common synchronous file of ci and qu is the LRC file, and the form of so-called LRC file to be exactly the temporal information back in simple terms following passage information.Wherein, the meaning of temporal information representative is exactly the zero-time of this section Word message in voice document.So,, also just can hear and the corresponding voice content of this section Word message as long as begin to play from this time of voice document.Also, just can produce many product or softwares on the market with ci and qu synchronizing function because there is the file of this form of similar LRC to occur.
But with present technology, the making major part of LRC file is to finish in the mode of manual manufacture.Just carry out the pairing time-tagging of sentence according to the content of literal and voice document.In simple terms, exactly with the time of voice document that word segment corresponds to, utilize artificial method to mark sentence by sentence.So, a large amount of time and the waste of manpower will be caused.
For example, TaiWan, China is applied for a patent " editing system of the vocal accompaniment lyrics and editor thereof and the method that shows " No. 92117564.But this patent provides and is applied on the computing machine executive's interface, edit the corresponding lyrics of vocal accompaniment music rhythm by the user, and when the zero-time that defines every section song is used to show, can show and change corresponding character according to the song duration accurately to present, the user can be followed easily sing.Its disclosed technology is to edit the corresponding lyrics of vocal accompaniment music rhythm by the user, just adopts the above-mentioned artificial mode of introducing that indicates the time voluntarily, and finishing accompanying song Chinese words file (lyrics) can the synchronous function of ci and qu.
In the at present relevant in addition research document, there is trial key vocabularies to be put in order and, realizes the keyword extraction of big glossary, and its discrimination power and identification usefulness are studied with fast algorithm with all key vocabularies structurings.And, be the speech control system of platform with PDA, the vector quantization process based on neural network is discussed, to the influence of voice system discrimination power.The method of using comprises utilizes Digital Signal Processing acquisition speech characteristic parameter, and the vector quantization method is done pre-treatment, and hidden Markov model is main identification and training algorithm.
Above-mentioned mentioned document, main research contents focuses on the skill of speech recognition.Can't reach the function of time that the corresponding text file of voice document is indicated automatically.Therefore, how allowing text file indicate the time automatically, and save time and the money that the artificial sign time is spent, is a problem demanding prompt solution.
Summary of the invention
In view of this present invention proposes the apparatus and method that a kind of text file indicates the time automatically, carries out text file by speech recognition and indicates the time automatically.Utilize the present invention each sentence in the text file can be indicated automatically time corresponding to voice document.Therefore, do not need again as conventional art, utilize artificial mode sentence by sentence the mark or character file correspond to time of voice document.So, with the cost of significantly saving time with manpower.
A kind of text file proposed by the invention indicates the device of time automatically and comprises: receiver module, voice identification module and sign module.
Receiver module receives text file and voice document.Wherein, text file is made up of a plurality of sentence.Voice identification module is converted to speech model with the sentence in the text file, and according to interval time voice document being divided into a plurality of frames (frame) and numbering in regular turn, calculates the match each other optimal voice path of (match) of frame and speech model.Indicate module and capture the numbering of the pairing frame of beginning of each sentence according to optimal voice path, by the numbering and the zero-time of the beginning that obtains each sentence interval time corresponding to voice document of frame, and the sign zero-time is in text file.
The present invention proposes a kind of text file and indicates time method automatically, carries out text file by speech recognition and indicates the time automatically, comprises the following step.Receive text file and voice document, and this literal file is made up of a plurality of sentence.Sentence in the conversion text file is a speech model.According to interval time voice document being divided into a plurality of frames (frame) and numbering in regular turn.Calculate the match each other optimal voice path of (match) of frame and speech model.Capture the numbering of the pairing frame of beginning of each sentence according to optimal voice path.According to the numbering of frame and the beginning that obtains each sentence interval time zero-time corresponding to voice document.At last, indicate zero-time in text file.
Each sentence in the text file can be indicated automatically zero-time by method of the present invention, not need for another example that conventional art equally utilizes artificial mode to indicate the time sentence by sentence, and then save a large amount of time and the cost of manpower corresponding to voice document.
Relevant preferred embodiment of the present invention and effect thereof, conjunction with figs. explanation as after.
Description of drawings
Fig. 1 indicates the schematic representation of apparatus of time automatically for text file of the present invention.
Fig. 2 is the synoptic diagram of voice identification module.
Fig. 3 is the optimal voice path synoptic diagram.
Fig. 4 indicates the time method process flow diagram automatically for text file of the present invention.
Fig. 5 is the thin portion method flow diagram of calculating optimum voice path.
Wherein, description of reference numerals is as follows:
S1, S2, S3, S4: sentence
F1～FN: frame
10: text file
12: voice document
14: speech model
20: receiver module
30: voice identification module
32: acquisition module
34: the first computing modules
36: the second computing modules
38: optimal voice path
40: indicate module
Embodiment
Fig. 1 indicates the schematic representation of apparatus of time automatically for text file of the present invention.As shown in Figure 1, text file of the present invention indicates the device of time automatically and comprises: receiver module 20, voice identification module 30 and indicate module 40.
Then according to interval time voice document 12 being divided into a plurality of frames (frame) and numbering in regular turn.Wherein, be about 23～30 milliseconds interval time.In hidden Markov model, the characteristic parameter that each frame presented can be regarded the output result under certain state as.And the transfer of state, and the output result under a state, can describe by probability model.Utilize hidden Markov model or other speech recognition notion no matter be, earlier voice document 12 is divided into basic phonetic unit, just so-called frame, carrying out follow-up speech recognition again handles, convenience and accuracy that speech recognition is handled can be improved, the speed in the computing can be accelerated simultaneously.
Next a plurality of frames of being divided according to voice document 12 of voice identification module 30 with the speech model that text file 10 is changed, calculate the match each other optimal voice path of (match) of frame and speech model.
Indicate module 40 according to the optimal voice path that voice identification module 30 is produced, capture the numbering of the pairing frame of beginning of each sentence in the text file 10.Numbering by frame and the beginning that obtains each sentence interval time are corresponding to the zero-time of voice document 12 again.Suppose, comprise 4 sentences with voice document 12 corresponding text files 10.When the frame zero-time of voice document 12 is 30 seconds, and be the beginning of second sentence of text file via the result of speech recognition, 30 seconds is the zero-time of second sentence in the text file 10 so.That is to say that when the reproduction time of voice document 12 was 30 seconds, the content of being play just was the beginning of second sentence in the text file 10,30 seconds was the zero-time that second sentence corresponds to voice document 12 in the text file 10 so.Same, when the frame zero-time of voice document 12 is 55 seconds, and be the beginning of the 3rd sentence of text file via the result of speech recognition, 55 seconds is the zero-time of the 3rd sentence in the text file 10 so.That is to say, when voice document 12 continues to play to the time when being 55 seconds, the content of being play just is the beginning of the 3rd sentence in the text file 10,55 seconds is the zero-time that the 3rd sentence corresponds to voice document 12 in the text file 10 so, by that analogy.
Wherein, capture according to optimal voice path each sentence in the text file 10 beginning after the numbering of corresponding frame, owing to can choose voluntarily according to user's demand or according to the requirement on calculating the interval time of frame.Therefore, the algorithm of the zero-time of each sentence can start pairing frame number by each sentence and be multiplied by the interval time of each frame and obtain.For example, supposing to be set at interval time 25 milliseconds and each frame does not all have overlappingly, and just voice document 12 every intervals just are divided into a frame for 25 milliseconds.Suppose, the beginning of second sentence in the text file 10, capturing corresponding frame number via optimal voice path is 1200, because the time that every frame comprised is 25 milliseconds, therefore the beginning of second sentence is that the numbering of frame is multiplied by interval time (1200*25ms=30s) corresponding to the zero-time of voice document 12 in the text file 10, and the beginning that can obtain second sentence is 30 seconds corresponding to the zero-time of voice document 12.Same, the beginning of the 3rd sentence in the text file 10, capturing corresponding frame number via optimal voice path is 2200, therefore the beginning of the 3rd sentence is that the numbering of frame is multiplied by interval time (2200*25ms=55s) corresponding to the zero-time of voice document 12, and the beginning that can obtain the 3rd sentence is 55 seconds corresponding to the zero-time of voice document 12.
At last, indicate module 40 and indicate zero-time in text file 10.After obtaining the zero-time of beginning corresponding to voice document 12 of each sentence in the text file 10, the zero-time of a sentence is shown in text file 10.So, similar LRC file is the same, and text file 10 not only writes down the word content corresponding to voice document 12, more writes down the zero-time of the beginning of each sentence.So voice document 12 needs only from the zero-time of some sentences and begins to play, and also just can hear and the corresponding voice content of this sentence word content, reaches the synchronous function of ci and qu.And do not need to utilize artificial mode to come the sign time like that in the conventional art for another example, each sentence in the text file 10 can be indicated automatically zero-time corresponding to voice document 12 by device of the present invention.
Please refer to Fig. 2, Fig. 2 is the synoptic diagram of voice identification module.The present invention's text file indicates in the device of time automatically, and voice identification module 30 comprises: acquisition module 32, first computing module 34 and second computing module 36.
Voice signal has an important characteristic, at different time, though the voice that send are that its waveform is not quite similar, and we can say that also voice are dynamic signals of a kind of time to time change with a word or same sound.Speech recognition will be found out regularity exactly from these dynamic signals, in case find after the regularity, how voice signal changes in time, on the whole can both point out their characteristic place, and then the voice signal identification is come out.This regularity is called characteristic parameter (featureparameter) in speech recognition, just can represent the parameter of voice signal characteristic.And the ultimate principle of speech recognition is exactly as the basis with these characteristic parameters.Therefore at the beginning, acquisition module 32 captures the pairing characteristic parameter of each frame in the voice document 12 earlier, in order to the processing of subsequent voice identification.
Because aforesaid speech model can belong to hidden Markov model, and hidden Markov model is the method for a kind of Probability ﹠ Statistics aspect, be fit to be used in the description of characteristics of speech sounds, because voice are random processing signals of a kind of multiparameter, via the processing of hidden Markov model, can infer accurately all parameters.Therefore, next first computing module 34 utilizes first algorithm computation to go out a comparison probability of each characteristic parameter and speech model.Wherein, first algorithm can be forward process (forwardprocedure) algorithm or reverse process (backward procedure) algorithm.The state number of supposing hidden Markov model is N, and hidden Markov model allows by a certain state transitions to any state, and then all state transition sequence numbers are N
TWhen the T value is too big, make that the calculated amount of probability can be too heavy.Therefore, can adopt forward process algorithm or reverse process algorithm, come accelerometer to calculate the comparison probability of characteristic parameter and speech model.
Fig. 3 is for being the optimal voice path synoptic diagram.As shown in Figure 3, the comparison probability that second computing module 36 is calculated according to first computing module 34, and utilize second algorithm, calculate optimal voice path 38.Wherein, second algorithm can adopt viterbi algorithm (viterbi algorithm).Suppose to have in the text file 10 four sentences to be in regular turn: S1, S2, S3 and S4.At first, preface be docile and obedient in these four sentences be converted to speech model 14, will be divided into a plurality of frames (F1～FN) again corresponding to the voice document 12 of text file 10.(F1～FN) is a transverse axis, and the speech model of being changed with text file 10 14 is that the longitudinal axis carries out identification and viterbi algorithm is with a plurality of frames of voice document 12.Viterbi algorithm can mark the most possible corresponding speech model of each frame.After the characteristic parameter of all frames in the voice document 12 is all handled, can obtain calculating: the optimal voice path 38 that whole frames are the most similar to speech model via viterbi algorithm.
As shown in Figure 3 as seen, by optimal voice path 38 can capture each sentence beginning the numbering of corresponding frame.According to the numbering of the frame of each sentence and the interval time that each frame is comprised, can obtain the zero-time of the beginning of each sentence corresponding to voice document 12.
Fig. 4 is for to indicate the time method process flow diagram automatically for text file, and Fig. 4 comprises the following step:
Step S10: receive text file and voice document.Wherein, text file and voice document are file in correspondence with each other, and text file is made up of a plurality of sentence.
Step S20: the sentence in the conversion text file is a speech model.Wherein, speech model belongs to hidden Markov model.
Step S30: the voice document that step S10 is received, according to being divided into a plurality of frames and numbering in regular turn interval time.Wherein, be about 23～30 milliseconds interval time.
Step S40: calculate the optimal voice path that frame and speech model match each other.This step can be subdivided into three steps again, will introduce below.
Step S50: capture according to optimal voice path each sentence beginning the numbering of corresponding frame.
Step S60: according to the numbering of frame and the beginning that obtains each sentence interval time zero-time corresponding to voice document.Because can choose voluntarily the interval time of frame according to user's demand or according to the requirement on calculating.Therefore, the algorithm of the zero-time of each sentence can start pairing frame number by each obtained sentence of step S50 and be multiplied by the interval time of each frame and obtain.
Step S70: last, the zero-time of beginning that indicates each sentence is in text file.So, text file more writes down the zero-time of the beginning of each sentence except the word content of record corresponding to voice document.Therefore, voice document if from the zero-time of some sentences begin to play, just can hear and the corresponding voice content of this sentence word content, and reach the synchronous function of ci and qu.Each sentence in the text file can be indicated automatically zero-time by method of the present invention, not need for another example that conventional art equally utilizes artificial mode to indicate the time sentence by sentence, and then save a large amount of time and the cost of manpower corresponding to voice document.
Above-mentioned steps S40 calculates the optimal voice path that frame and speech model match each other, and comprises the following step, please refer to Fig. 5, and Fig. 5 is the thin portion method flow diagram of calculating optimum voice path.
Step S42: capture the corresponding characteristic parameter of each frame.Though voice signal is a kind of Dynamic Signal of time to time change, as long as but find out in the voice signal each in short-term apart from (short time) or be called the regularity of a frame, no matter how voice signal changes in time so, can both find out its characteristic place haply, and then the voice signal identification is come out.And this regularity is called characteristic parameter (feature parameter) in speech recognition, just can represent the parameter of voice signal characteristic.Therefore, the characteristic parameter with each frame captures out earlier, in order to the processing of subsequent voice identification.
Step S44: utilize first algorithm computation to go out the comparison probability of each characteristic parameter and speech model.Wherein, first algorithm can be forward process algorithm or reverse process algorithm.
Step S46: each characteristic parameter that is calculated according to step S44 and the comparison probability of speech model, utilize second algorithm computation to go out optimal voice path again.Wherein, second algorithm can adopt viterbi algorithm.As shown in Figure 3, the optimal voice path of utilizing viterbi algorithm to calculate captures the numbering of the pairing frame of beginning of each sentence in the text file again by optimal voice path.According to the numbering of the frame of each sentence and the interval time that each frame is comprised, can obtain the zero-time of the beginning of each sentence corresponding to voice document.
Though technology contents of the present invention with preferred embodiment openly as above; yet it is not in order to limit scope of patent protection of the present invention; any those of ordinary skill in the art; do not breaking away from change and the retouching that spirit of the present invention is done; all should be covered by in the scope of patent protection of the present invention, so protection scope of the present invention should be as the criterion with the scope of patent protection that affiliated claims define.
Claims (16)
1. a text file indicates the device of time automatically, comprising:
Receiver module receives text file and voice document, and this literal file is made up of a plurality of sentence;
Voice identification module is converted to speech model with the described sentence in this literal file, and according to an interval time this voice document being divided into a plurality of frames and numbering in regular turn, calculates the optimal voice path that described frame and this speech model match each other; And
Indicate module, capture the numbering of pairing this frame of beginning of each described sentence according to this optimal voice path, by the numbering of this frame and obtain the initial time of the beginning of each described sentence this interval time, and indicate this zero-time in this literal file corresponding to this voice document.
2. text file as claimed in claim 1 indicates the device of time automatically, and wherein this speech model belongs to hidden Markov model.
3. text file as claimed in claim 1 indicates the device of time automatically, wherein is about 23～30 milliseconds this interval time.
4. text file as claimed in claim 1 indicates the device of time automatically, and wherein this voice identification module also comprises:
Acquisition module captures the corresponding characteristic parameter of each described frame;
First computing module utilizes first algorithm computation to go out the comparison probability of each described characteristic parameter and this speech model; And
Second computing module according to this comparison probability and utilize second algorithm, calculates this optimal voice path.
5. text file as claimed in claim 4 indicates the device of time automatically, and wherein this first algorithm is the forward process algorithm.
6. text file as claimed in claim 4 indicates the device of time automatically, and wherein this first algorithm is the reverse process algorithm.
7. text file as claimed in claim 4 indicates the device of time automatically, and wherein this second algorithm is a viterbi algorithm.
8. text file as claimed in claim 1 indicates the device of time automatically, and wherein this zero-time is multiplied by this interval time by the numbering of this frame and obtains.
9. a text file indicates time method automatically, comprises the following steps:
Receive text file and voice document, this literal file is made up of a plurality of sentence;
The described sentence of changing in this literal file is a speech model;
According to interval time this voice document being divided into a plurality of frames and numbering in regular turn;
Calculate described frame and this speech model optimal voice path that one of matches each other;
Capture the numbering of pairing this frame of beginning of each described sentence according to described optimal voice path;
According to the numbering of this frame and obtain the initial time of the beginning of each described sentence this interval time corresponding to this voice document; And
Indicate this zero-time in this literal file.
10. text file as claimed in claim 9 indicates time method automatically, and wherein this speech model belongs to a hidden Markov model.
11. text file as claimed in claim 9 indicates time method automatically, wherein is about 23～30 milliseconds this interval time.
12. text file as claimed in claim 9 indicates time method automatically, wherein this calculation procedure also comprises the following steps:
Capture the corresponding characteristic parameter of each described frame;
Utilize one first algorithm computation to go out one of each described characteristic parameter and this speech model comparison probability; And
According to this comparison probability and utilize second algorithm, calculate this optimal voice path.
13. text file as claimed in claim 12 indicates time method automatically, wherein this first algorithm is the forward process algorithm.
14. text file as claimed in claim 12 indicates time method automatically, wherein this first algorithm is the reverse process algorithm.
15. text file as claimed in claim 12 indicates time method automatically, wherein this second algorithm is a viterbi algorithm.
16. text file as claimed in claim 9 indicates time method automatically, wherein this zero-time is multiplied by this interval time by the numbering of this frame and obtains.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007100886277A CN101266790A (en) | 2007-03-16 | 2007-03-16 | Device and method for automatic time marking of text file |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007100886277A CN101266790A (en) | 2007-03-16 | 2007-03-16 | Device and method for automatic time marking of text file |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101266790A true CN101266790A (en) | 2008-09-17 |
Family
ID=39989144
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007100886277A Pending CN101266790A (en) | 2007-03-16 | 2007-03-16 | Device and method for automatic time marking of text file |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101266790A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109217965A (en) * | 2018-09-26 | 2019-01-15 | 杭州当虹科技股份有限公司 | A kind of SDIOverIP reception inter-system synchronization method based on timestamp |
| CN109920406A (en) * | 2019-03-28 | 2019-06-21 | 国家计算机网络与信息安全管理中心 | A kind of dynamic voice recognition methods and system based on variable initial position |
| CN112102847A (en) * | 2020-09-09 | 2020-12-18 | 四川大学 | Audio and slide content alignment method |
-
2007
- 2007-03-16 CN CNA2007100886277A patent/CN101266790A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109217965A (en) * | 2018-09-26 | 2019-01-15 | 杭州当虹科技股份有限公司 | A kind of SDIOverIP reception inter-system synchronization method based on timestamp |
| CN109920406A (en) * | 2019-03-28 | 2019-06-21 | 国家计算机网络与信息安全管理中心 | A kind of dynamic voice recognition methods and system based on variable initial position |
| CN109920406B (en) * | 2019-03-28 | 2021-12-03 | 国家计算机网络与信息安全管理中心 | Dynamic voice recognition method and system based on variable initial position |
| CN112102847A (en) * | 2020-09-09 | 2020-12-18 | 四川大学 | Audio and slide content alignment method |
| CN112102847B (en) * | 2020-09-09 | 2022-08-09 | 四川大学 | Audio and slide content alignment method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110148427B (en) | Audio processing method, device, system, storage medium, terminal and server | |
| CN101399036B (en) | Device and method for conversing voice to be rap music | |
| CN109285537B (en) | Acoustic model establishing method, acoustic model establishing device, acoustic model synthesizing method, acoustic model synthesizing device, acoustic model synthesizing equipment and storage medium | |
| CN108847215B (en) | Method and device for voice synthesis based on user timbre | |
| CN101739870A (en) | Interactive language learning system and method | |
| CN103503015A (en) | System for creating musical content using a client terminal | |
| Fujihara et al. | Lyrics-to-audio alignment and its application | |
| CN110675886A (en) | Audio signal processing method, audio signal processing device, electronic equipment and storage medium | |
| CN110691258A (en) | Program material manufacturing method and device, computer storage medium and electronic equipment | |
| CN112289300B (en) | Audio processing method and device, electronic equipment and computer readable storage medium | |
| CN111739536A (en) | Audio processing method and device | |
| Gu et al. | Mm-alt: A multimodal automatic lyric transcription system | |
| Lee et al. | Word level lyrics-audio synchronization using separated vocals | |
| CN101266790A (en) | Device and method for automatic time marking of text file | |
| CN112750421B (en) | Singing voice synthesis method and device and readable storage medium | |
| US20080189105A1 (en) | Apparatus And Method For Automatically Indicating Time in Text File | |
| Xu et al. | Automatic music summarization based on temporal, spectral and cepstral features | |
| Kruspe et al. | Retrieval of Textual Song Lyrics from Sung Inputs. | |
| CN1153127C (en) | Intelligent common spoken Chinese phonetic input method and dictation machine | |
| Schuller et al. | Incremental acoustic valence recognition: an inter-corpus perspective on features, matching, and performance in a gating paradigm | |
| Hacioglu et al. | Parsing speech into articulatory events | |
| Zhu | Multimedia recognition of piano music based on the hidden markov model | |
| Cen et al. | Segmentation of speech signals in template-based speech to singing conversion | |
| CN1182259A (en) | Pronunciation training system and method | |
| Iriondo et al. | Objective and subjective evaluation of an expressive speech corpus |
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 |
Open date: 20080917 |