CN107910004A - Voiced translation processing method and processing device - Google Patents
Voiced translation processing method and processing device Download PDFInfo
- Publication number
- CN107910004A CN107910004A CN201711107221.9A CN201711107221A CN107910004A CN 107910004 A CN107910004 A CN 107910004A CN 201711107221 A CN201711107221 A CN 201711107221A CN 107910004 A CN107910004 A CN 107910004A
- Authority
- CN
- China
- Prior art keywords
- synthetic speech
- voice signal
- speech signal
- signal
- text data
- 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
- 238000013519 translation Methods 0.000 title claims abstract description 80
- 238000012545 processing Methods 0.000 title claims abstract description 16
- 238000003672 processing method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 55
- 238000002156 mixing Methods 0.000 claims abstract description 30
- 230000006854 communication Effects 0.000 claims abstract description 17
- 238000004891 communication Methods 0.000 claims abstract description 15
- 230000015572 biosynthetic process Effects 0.000 claims description 56
- 238000003786 synthesis reaction Methods 0.000 claims description 56
- 230000001755 vocal effect Effects 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 7
- 230000001052 transient effect Effects 0.000 claims description 6
- 230000005055 memory storage Effects 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 26
- 238000010586 diagram Methods 0.000 description 5
- 230000005236 sound signal Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 239000003086 colorant Substances 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 230000009469 supplementation Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- NGVDGCNFYWLIFO-UHFFFAOYSA-N pyridoxal 5'-phosphate Chemical compound CC1=NC=C(COP(O)(O)=O)C(C=O)=C1O NGVDGCNFYWLIFO-UHFFFAOYSA-N 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000007306 turnover Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/40—Processing or translation of natural language
- G06F40/58—Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L13/00—Speech synthesis; Text to speech systems
- G10L13/02—Methods for producing synthetic speech; Speech synthesisers
- G10L13/04—Details of speech synthesis systems, e.g. synthesiser structure or memory management
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Acoustics & Sound (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
Abstract
The embodiment of the present invention provides a kind of voiced translation processing method and processing device, belongs to language processing techniques field.This method includes：During the first synthetic speech signal is reported, if receiving the mixing voice signal for including the first synthetic speech signal of part, stop reporting the first synthetic speech signal.The first synthetic speech signal of part is filtered out from mixing voice signal, obtains current round voice signal to be translated, and be used as targeted voice signal.Based on targeted voice signal, the second synthetic speech signal is obtained, and report the second synthetic speech signal.The embodiment of the present invention stops reporting the first synthetic speech signal, and report the second synthetic speech signal when receiving the mixing voice signal for including the first synthetic speech signal of part.Due to any one party in communication process, report process can be interrupted at any time according to full-duplex mode, and terminated without waiting until that a wheel reports process every time, so as to improve communication efficiency.
Description
Technical field
The present embodiments relate to language processing techniques field, more particularly, to a kind of voiced translation processing method and
Device.
Background technology
At present, language communication becomes the important topic that different racial groups are faced in mutual exchange.For example,
In double or multi-person conference, voiced translation can be realized by automatic speech translation system.Wherein, automatic speech translation system is usual
It is made of speech recognition, machine translation and phonetic synthesis three parts.Source languages voice signal obtains source languages by speech recognition
Text data, then translates into target language text data, finally by mesh by machine translation by source languages text data
Poster kind text data carries out phonetic synthesis, obtains target language voice signal and is reported.When carrying out voiced translation, need
After waiting until that the target language voice signal of last round of time has been reported, the source languages voice signal of next one can be just translated.
Due in instant communication exchange, listener may just can be bright when carrying out half for the report process of last round of time
White speaker to be said if implication, and listener need to continue after waiting last round of secondary report process it is next
The communication of round, so that communication efficiency is relatively low.
The content of the invention
To solve the above-mentioned problems, the embodiment of the present invention provides one kind and overcomes the above problem or solve at least in part
State the voiced translation processing method and processing device of problem.
First aspect according to embodiments of the present invention, there is provided a kind of voiced translation processing method, this method include：
During being reported to the first synthetic speech signal, voice letter is synthesized if receiving and including part first
Number mixing voice signal, then stop reporting the first synthetic speech signal, the first synthetic speech signal is time to turn over via last round of
Translate and phonetic synthesis after it is obtained；
The first synthetic speech signal of part is filtered out from mixing voice signal, obtains current round voice letter to be translated
Number, and it is used as targeted voice signal；
Based on targeted voice signal, the second synthetic speech signal is obtained, and reports the second synthetic speech signal, the second synthesis
Voice signal be targeted voice signal is translated and phonetic synthesis after it is obtained.
Method provided in an embodiment of the present invention, by during being reported to the first synthetic speech signal, if connecing
The mixing voice signal for including the first synthetic speech signal of part is received, then stops reporting the first synthetic speech signal.From mixed
Close filtered speech signal and fall the first synthetic speech signal of part, obtain current round voice signal to be translated, and be used as target
Voice signal.Based on targeted voice signal, the second synthetic speech signal is obtained, and report the second synthetic speech signal.Due to ditch
Any one party during logical, can interrupt report process at any time according to full-duplex mode, and without waiting until a carousel every time
Report process terminates so that while communication efficiency is improved, also may be such that link up between different language user it is more natural and tripping.
The possible implementation of with reference to first aspect the first, in second of possible implementation, the first synthesis
Voice signal and targeted voice signal be identical languages type or the first synthetic speech signal with targeted voice signal for not
Same languages type.
The possible implementation of with reference to first aspect the first, in the third possible implementation, reports second
Before synthetic speech signal, further include：
The identification text data to being obtained after targeted voice signal progress speech recognition is obtained, is obtained to identifying text data
The target text data obtained after being translated, and phonetic synthesis is carried out to target text data, obtain the second synthesis voice letter
Number.
The third possible implementation with reference to first aspect, in the 4th kind of possible implementation, obtains to knowing
The target text data that other text data obtains after being translated, including：
Determine the corresponding source languages type of identification text data, and determine that source languages type corresponds to according to default correspondence
Target language type；
Target language type and identification text data are inputted to translation coding/decoding model, export target text data.
The 4th kind of possible implementation with reference to first aspect, in the 5th kind of possible implementation, based on target
Vocal print feature in voice signal, determines the corresponding default languages type of vocal print feature, to be used as the corresponding mesh of source languages type
Poster type.
The third possible implementation with reference to first aspect, in the 6th kind of possible implementation, obtains to knowing
The target text data that other text data obtains after being translated, including：
If judgement knows that targeted voice signal and the information that the first synthetic speech signal transmits respectively are interrelated, it is based on
The voice signal and/or translation result of round before current round, translate identification text data, to obtain target text
Data.
The third possible implementation with reference to first aspect, in the 7th kind of possible implementation, to target text
Notebook data carries out phonetic synthesis, obtains the second synthetic speech signal, including：
Voice broadcast parameter is obtained, target text data and voice broadcast parameter are inputted to phonetic synthesis model, output
Second synthetic speech signal, used tone color ginseng when reporting the second synthetic speech signal is included at least in voice broadcast parameter
Number.
Second aspect according to embodiments of the present invention, there is provided a kind of voiced translation processing unit, the device include：
Stop broadcasting module, for during being reported to the first synthetic speech signal, including if receiving
The mixing voice signal of the first synthetic speech signal of part, then stop reporting the first synthetic speech signal, the first synthesis voice letter
Number it is via obtained after last round of time translation and phonetic synthesis；
Filtering module, for filtering out the first synthetic speech signal of part from mixing voice signal, obtains current round and treats
The voice signal of translation, and it is used as targeted voice signal；
Broadcasting module, for based on targeted voice signal, obtaining the second synthetic speech signal, and report the second synthesis voice
Signal, the second synthetic speech signal be targeted voice signal is translated and phonetic synthesis after it is obtained.
The third aspect according to embodiments of the present invention, there is provided a kind of voiced translation processing equipment, including：
At least one processor；And
At least one processor being connected with processor communication, wherein：
Memory storage has the programmed instruction that can be executed by processor, and the instruction of processor caller is able to carry out first party
The voiced translation processing method that any possible implementation is provided in the various possible implementations in face.
According to the fourth aspect of the invention, there is provided a kind of non-transient computer readable storage medium storing program for executing, non-transient computer
Readable storage medium storing program for executing stores computer instruction, and computer instruction makes the various possible implementations of computer execution first aspect
In the voiced translation processing method that is provided of any possible implementation.
It should be appreciated that the general description and following detailed description of the above are exemplary and explanatory, can not
Limit the embodiment of the present invention.
Brief description of the drawings
Fig. 1 is a kind of flow diagram of voiced translation processing method of the embodiment of the present invention；
Fig. 2 is a kind of flow diagram of voiced translation processing method of the embodiment of the present invention；
Fig. 3 is a kind of flow diagram of voiced translation processing method of the embodiment of the present invention；
Fig. 4 is a kind of block diagram of voiced translation processing unit of the embodiment of the present invention；
Fig. 5 is a kind of block diagram of voiced translation processing equipment of the embodiment of the present invention.
Embodiment
With reference to the accompanying drawings and examples, the embodiment of the embodiment of the present invention is described in further detail.With
Lower embodiment is used to illustrate the embodiment of the present invention, but is not limited to the scope of the embodiment of the present invention.
At present, the people of different language realizes when carrying out communication exchange typically by automatic speech translation system.Its
In, automatic speech translation system is usually made of speech recognition, machine translation and phonetic synthesis three parts.Source languages voice signal
Source languages text data is obtained by speech recognition, source languages text data is then translated into by target language by machine translation
Text data, carries out phonetic synthesis finally by target language text data, obtains target language voice signal and broadcast
Report.When carrying out voiced translation, it is necessary to after the target language voice signal of last round of time has been reported, can just carry out next
Translation, phonetic synthesis and the report of round.
For example, communication exchange between user A and user B, user A speaks English, and user B says Chinese.User A says an English
Language, the process by translation and phonetic synthesis obtains a Chinese sentence, and is reported.Complete after whole sentence reports, user A can be after
It is continuous to say an English or a Chinese sentence is said by user B, and repeat the process of above-mentioned translation, phonetic synthesis and report.That is,
User A and user B need that after system is reported, new voice data could be received, and translated, phonetic synthesis and
Report.
After a voice data having been said in view of user, it may be necessary to which the voice data just said is supplemented or changed.
In addition, when reporting synthetic speech signal, the user of report is listened to may not be needed to hear out just it will be appreciated that the intention of spoken user.
For said circumstances, reported if completing the whole sentence of last round of time according to above-mentioned flow, then carry out the translation of next one, voice conjunction
Into and report, then can compare expend the time.In view of the above-mentioned problems, an embodiment of the present invention provides a kind of voiced translation processing side
Method.This method can be applied to voice collecting, translation, synthesis and report function terminal or system, can be applied to two people or
The more people of person link up scene, and the embodiment of the present invention is not especially limited this.Referring to Fig. 1, this method includes：101st, closed to first
During being reported into voice signal, if receiving the mixing voice signal for including the first synthetic speech signal of part,
Then stop reporting the first synthetic speech signal, the first synthetic speech signal is via institute after last round of translation and phonetic synthesis
Obtain；102nd, the first synthetic speech signal of part is filtered out from mixing voice signal, obtains current round voice to be translated
Signal, and it is used as targeted voice signal；103rd, based on targeted voice signal, the second synthetic speech signal is obtained, and report second
Synthetic speech signal, the second synthetic speech signal be targeted voice signal is translated and phonetic synthesis after it is obtained.
In above-mentioned steps 101, the first synthetic speech signal for last round of time source languages signal by collection, translation and
Obtained by after phonetic synthesis., can be while listening for whether having new source during being reported to the first synthetic speech signal
Languages voice signal, namely monitor whether have user said again needs translate report if.Specifically, can be by opening a prison
Thread is listened to monitor, the embodiment of the present invention is not especially limited this.In snoop procedure, due to the first synthesis language of last round of time
Sound signal is also being reported, so that the voice signal listened to is except including new source languages voice signal, (i.e. user is new
Speech) outside, also there are the first synthetic speech signal of part, so that what is received includes the first synthetic speech signal of part
Mixing voice signal.After the mixing voice signal for including the first synthetic speech signal of part is received, illustrate useful at this time
Family made a statement, which is probably that the user of last round of speech requires supplementation with or changes, so as to interrupt the first synthesis voice letter
Number report process make a speech.Or, it is also possible to the user for listening report, which does not hear out, has just understood last round of speech
The intention that user is passed on, so that the report process for interrupting the first synthetic speech signal is made a speech.
Due in addition to including the mixing voice signal of the first synthetic speech signal of part, being gone back in mixing voice signal
Comprising current round voice signal to be translated, and the voice signal to be translated to current round is subsequently needed to be translated, language
Sound is synthesized and reported, so that, it is necessary to which part first is filtered out from mixing voice signal synthesizes voice letter in above-mentioned steps 102
Number, so as to obtain current round voice signal to be translated.The embodiment of the present invention from mixing voice signal to not filtering out portion
The mode of the first synthetic speech signal is divided specifically to limit, from creolized language message including but not limited to by way of echo cancellor
The first synthetic speech signal of part is filtered out in number.Wherein, the calculating process of echo cancellor can be as follows：
By taking audiomonitor is microphone as an example, it is assumed that the first synthetic speech signal of part of report is s (t), m-th of Mike
The channel transfer function that wind receives is hm(t), the voice signal to be translated that user newly inputs is xm(t), then microphone receives
Observation signal ym(t), as shown in following equation：
ym(t)=s (t) * hm(t)+xm(t)
As the voice signal x to be translated not inputted newlym(t) when, channel transfer function h can be estimated in advancem(t).When
There is the voice signal x to be translated newly inputtedm(t) when, echo cancellor can be carried out to mixing voice signal.Due to ym(t), s (t),
hm(t) it is known that so as to which current round voice signal to be translated is calculated by equation below, and it is used as target voice
Signal, specific formula are as follows：
x′m(t)=y (t)-s (t) * hm(t)
After targeted voice signal is obtained, targeted voice signal can be translated and phonetic synthesis, to obtain working as front-wheel
The second secondary synthetic speech signal, and report the second synthetic speech signal.
Method provided in an embodiment of the present invention, by during being reported to the first synthetic speech signal, if connecing
The mixing voice signal for including the first synthetic speech signal of part is received, then stops reporting the first synthetic speech signal.From mixed
Close filtered speech signal and fall the first synthetic speech signal of part, obtain current round voice signal to be translated, and be used as target
Voice signal.Based on targeted voice signal, the second synthetic speech signal is obtained, and report the second synthetic speech signal.Due to ditch
Any one party during logical, can interrupt report process at any time according to full-duplex mode, and without waiting until a carousel every time
Report process terminates so that while communication efficiency is improved, also may be such that link up between different language user it is more natural and tripping.
Content based on above-described embodiment, any one party in communication process can interrupt report process according to demand, from
And the user to talk can interrupt the report process of oneself speech, the user for listening report can interrupt last round of secondary user's speech
Report process.Accordingly, as a kind of alternative embodiment, the first synthetic speech signal is identical languages with targeted voice signal
Type or the first synthetic speech signal are different languages types from targeted voice signal.
Wherein, the voice signal before the translation of the first synthetic speech signal is identical languages type with targeted voice signal
When, then the report process of explanation last round of oneself speech that be probably the User break of last round of time speech, or can also be
The user of current round speech is identical with the languages type used by a user of last round of speech, and the use of current round speech
The report process of last round of secondary user's speech is interrupted in family.When the first synthetic speech signal from targeted voice signal is different languages
During type, then it is probably the last round of secondary user's speech of User break for listening report that explanation, which is, report process, or may be used also
The report process for last round of oneself speech that can be the User break of last round of time speech, and used different from last round of time
Languages type speech.Based on the above, the embodiment of the present invention is applicable to the ditch of different language between double or more people
Logical scene, such as double or multi-person conference.
Content based on above-described embodiment, it is necessary to obtain the second synthesis voice before the second synthetic speech signal is reported
Signal.Accordingly, as a kind of alternative embodiment, an embodiment of the present invention provides a kind of method for obtaining synthetic speech signal.
Referring to Fig. 2, this method includes：201st, the identification text data to being obtained after targeted voice signal progress speech recognition is obtained；
202nd, the target text data obtained after being translated to identification text data are obtained；203rd, voice is carried out to target text data
Synthesis, obtains the second synthetic speech signal.
Content based on above-described embodiment, as a kind of alternative embodiment, the embodiment of the present invention additionally provides a kind of acquisition
The method of target text data.Referring to Fig. 3, this method includes：2011st, the corresponding source languages type of identification text data is determined,
And determine the corresponding target language type of source languages type according to default correspondence；2012nd, by target language type and identification
Text data is inputted to translation coding/decoding model, exports target text data.
In above-mentioned steps 2011, the corresponding source languages type of identification text data need to be first determined.The embodiment of the present invention is not
Mode to determining the corresponding source languages type of identification text data specifically limits, and includes but not limited to the following two kinds mode.
First way：Acoustic feature based on targeted voice signal determines.
Specifically, the acoustic feature of targeted voice signal, such as spectrum signature be can extract：Mel cepstrum coefficients (Mel-
Frequency Cepstral Coefficients, MFCC), perceive linear predictor coefficient (Linear Predictive
Coding, PLP) etc., acoustic feature is inputted to languages identification model, languages prediction is carried out to targeted voice signal.Languages are known
The output result of other model is the probability that targeted voice signal is each languages type, and the languages of select probability maximum are as mesh
The corresponding languages of poster sound signal, namely the corresponding source languages type of identification text data is determined.Wherein, languages identification model
Common classification model generally in pattern-recognition, specifically can extract every voice letter by collecting a large amount of voice signals in advance
Number acoustic feature, mark every voice signal languages type build to obtain.
The second way, the recognition result based on targeted voice signal determine.
Specifically, the corresponding speech recognition modeling of each languages being currently related to is utilized respectively to carry out targeted voice signal
Speech recognition, obtains targeted voice signal and corresponds to the identification text data of each languages and corresponding recognition confidence, selection is known
The identification text data of other confidence level maximum corresponds to languages of the languages as targeted voice signal.Wherein, speech recognition process one
As be：End-point detection first is carried out to targeted voice signal, obtains the starting point and end point of efficient voice section.Then endpoint is examined
The efficient voice section measured carries out feature extraction, recycles the characteristic of extraction and trained in advance acoustic model and language
Model is decoded, and obtains the confidence level that current speech data corresponds to identification text and corresponding identification text.
For example, it is assumed that targeted voice signal corresponds to languages as Chinese；The languages being currently related to are Chinese and English.To target
When voice signal carries out languages identification, Chinese speech recognition model and English speech recognition modeling are utilized respectively to target language message
Number speech recognition is carried out, obtain the corresponding Chinese identification text data and corresponding recognition confidence 0.9 of targeted voice signal, English
Identify text data and corresponding recognition confidence 0.2.The larger identification text data of selection recognition confidence corresponds to languages, i.e., in
Text corresponds to languages as targeted voice signal.Further, the identification that each languages can also be corresponded to identification text data is put
Reliability and language model scores are merged, and the languages corresponding to the maximum identification text data of selection fusion score are as target
The corresponding languages of voice signal.Wherein, fusion method can be linear weighting method, and the embodiment of the present invention does not limit this specifically
It is fixed.
After the corresponding source languages type of identification text data is determined, source languages type can be determined according to default correspondence
Corresponding target language type.Wherein, languages type in source is languages used in spoken user, and target language type is generally
Other languages types in the languages that active user's exchange is related in addition to the languages type of source.Wherein, target language type can be with
It is corresponding respectively to carry out communication exchange using the user of two languages and using the user of multiple languages for one or more.
For example, the languages that are related to are Chinese, English, French, Thai, Hindi, German etc. when if active user exchanges,
Chinese speech signal can be predefined and translate into English, French bilingual every time, that is, using English and French as mesh
Poster type.
In addition to pre-setting fixed target language type, the preference or demand of spoken user are also based on
To determine the corresponding target language type of source languages type.It is not right accordingly, as a kind of alternative embodiment, the embodiment of the present invention
Specifically limited in the way of default correspondence determines the corresponding target language type of source languages type, including it is but unlimited
In：Based on the vocal print feature in targeted voice signal, the corresponding default languages type of vocal print feature is determined, to be used as source languages class
The corresponding target language type of type.
If for example, meeting in leader speech need to be automatically translated into English, target language when can leader be talked in advance
Type is determined as English, and establishes correspondence between vocal print feature when talking with leader, so that in meeting is extracted
After leading the vocal print feature of speech, it is English that can directly determine target language type.
In above-mentioned steps 2012, inputted by target language type and identification text data to translation coding/decoding model,
When exporting target text data, the translation coding/decoding model based on neutral net can be used, identification text data is translated into
Corresponding target language text data, the embodiment of the present invention are not especially limited this.Wherein, each languages type is corresponding turns over
Translating coding/decoding model can build to obtain using a large amount of training datas in advance.
In addition to being translated above by translation coding/decoding model to identification text data, it is contemplated that above and below during speech
Relevance between literary content, the embodiment of the present invention additionally provide a kind of method for obtaining target text data, including but unlimited
In：If judgement knows that targeted voice signal and the information that the first synthetic speech signal transmits respectively are interrelated, based on current
The voice signal and/or translation result of round before round, translate identification text data, to obtain target text number
According to.
Specifically, it can determine whether that the targeted voice signal of current round passes respectively with first synthetic speech signal of last round of time
Whether the information passed is interrelated, if both related (such as during report, the User break of speech, which is reported, to be required supplementation with), then
After the targeted voice signal of current round can be merged with last round of translated voice signal in translation again
Translation, finally reports out by translation result again；The translation result of two-wheeled voice signal can also be merged upon translation
Afterwards, then report out；It can be combined with last round of translated voice signal, the translation result of last round of time, target language
The translation result of sound signal and targeted voice signal, reports out again after two-wheeled translation result is merged；It is in addition, above-mentioned
During identification text data to current round is translated, except the voice signal of last round of time and/or translation are tied
Fruit as reference outside, can also be by the voice signal of preceding n rounds and/or translation result as reference, the embodiment of the present invention pair
This is not especially limited.Wherein, n is more than or equal to 1.
After target text data are got, phonetic synthesis can be carried out to target text data, obtain the second synthesis voice
Signal.Accordingly, as a kind of alternative embodiment, on carrying out phonetic synthesis to target text data, the second synthesis language is obtained
The mode of sound signal, the embodiment of the present invention are not especially limited this, include but not limited to：Voice broadcast parameter is obtained, by mesh
Mark text data and voice broadcast parameter are inputted to phonetic synthesis model, export the second synthetic speech signal, voice broadcast parameter
In include at least report the second synthetic speech signal when used tamber parameter.
Wherein, when carrying out phonetic synthesis, a fixed speaker model can be selected to be synthesized, can such as uses one
A sound is neutral, the synthetic model of simple and honest sound synthesizes corresponding second synthetic speech signal.It is, of course, also possible to select individual character
The speaker model of change is synthesized.Specifically, the sound of a variety of different tone colors can be included in speech translation system, user can be with
Oneself selection, can also be made choice, the embodiment of the present invention does not make this to have by system according to the user information of active user
Body limits.Wherein, user information includes but not limited to the gender of user, age, tone color etc..For example, if the user for listening report is
Male, system can automatically select female speakers model, to synthesize the second synthetic speech signal of women sounding.
It is, of course, also possible to changed using sound, by the sound of synthesis be converted into user's tone color similar in sound broadcast
Report.For example, after being translated to the voice signal to be translated that user A is inputted, changed using sound so that transformed language
The tone color of sound signal is close with user's A tone colors, and transformed voice signal is reported out.
Method provided in an embodiment of the present invention, by during being reported to the first synthetic speech signal, if connecing
The mixing voice signal for including the first synthetic speech signal of part is received, then stops reporting the first synthetic speech signal.From mixed
Close filtered speech signal and fall the first synthetic speech signal of part, obtain current round voice signal to be translated, and be used as target
Voice signal.Based on targeted voice signal, the second synthetic speech signal is obtained, and report the second synthetic speech signal.Due to ditch
Any one party during logical, can interrupt report process at any time according to full-duplex mode, and without waiting until a carousel every time
Report process terminates so that while communication efficiency is improved, also may be such that link up between different language user it is more natural and tripping.
Further, since voice signal and/or translation result based on round before current round, to identification text data into
Row translation, the relevance of context during so as to make full use of speech, improves translation accuracy rate.
Finally, due to which languages type corresponding target language type in source can be determined according to default correspondence, can such as be based on
Vocal print feature in targeted voice signal, determines target language type, so as to meet preference and demand of the user in translation,
Realize personalized customization.
It should be noted that above-mentioned all alternative embodiments, can use any combination to form the optional implementation of the present invention
Example, this is no longer going to repeat them.
Content based on above-described embodiment, an embodiment of the present invention provides a kind of voiced translation processing unit, which turns over
Processing unit is translated to be used to perform the voiced translation processing method in above method embodiment.Referring to Fig. 4, which includes：
Stop broadcasting module 401, for during being reported to the first synthetic speech signal, being included if receiving
There is the mixing voice signal of the first synthetic speech signal of part, then stop reporting the first synthetic speech signal, the first synthesis voice
Signal is via obtained after last round of translation and phonetic synthesis；
Filtering module 402, for filtering out the first synthetic speech signal of part from mixing voice signal, obtains current round
Voice signal to be translated, and it is used as targeted voice signal；
Broadcasting module 403, for based on targeted voice signal, obtaining the second synthetic speech signal, and report the second synthesis
Voice signal, the second synthetic speech signal be targeted voice signal is translated and phonetic synthesis after it is obtained.
As a kind of alternative embodiment, the first synthetic speech signal and targeted voice signal for identical languages type or
First synthetic speech signal is different languages types from targeted voice signal.
As a kind of alternative embodiment, which further includes：
First acquisition module, for obtaining the identification text data to being obtained after targeted voice signal progress speech recognition；
Second acquisition module, for obtaining the target text data obtained to identifying text data after translating；
Voice synthetic module, for carrying out phonetic synthesis to target text data, obtains the second synthetic speech signal.
As a kind of alternative embodiment, the second acquisition module, including：
Determination unit, for determining the corresponding source languages type of identification text data, and determines according to default correspondence
The corresponding target language type of source languages type；
Translation unit, for inputting target language type and identification text data to translation coding/decoding model, exports mesh
Mark text data.
As a kind of alternative embodiment, determination unit, for based on the vocal print feature in targeted voice signal, determining vocal print
The corresponding default languages type of feature, to be used as the corresponding target language type of source languages type.
As a kind of alternative embodiment, the second acquisition module, for knowing targeted voice signal and the first synthesis when judgement
When the information that voice signal transmits respectively is interrelated, then voice signal and/or translation knot based on round before current round
Fruit, translates identification text data, to obtain target text data.
As a kind of alternative embodiment, voice synthetic module, for obtaining voice broadcast parameter, by target text data and
Voice broadcast parameter is inputted to phonetic synthesis model, is exported the second synthetic speech signal, is included at least and broadcast in voice broadcast parameter
Used tamber parameter when reporting the second synthetic speech signal.
Device provided in an embodiment of the present invention, by during being reported to the first synthetic speech signal, if connecing
The mixing voice signal for including the first synthetic speech signal of part is received, then stops reporting the first synthetic speech signal.From mixed
Close filtered speech signal and fall the first synthetic speech signal of part, obtain current round voice signal to be translated, and be used as target
Voice signal.Based on targeted voice signal, the second synthetic speech signal is obtained, and report the second synthetic speech signal.Due to ditch
Any one party during logical, can interrupt report process at any time according to full-duplex mode, and without waiting until a carousel every time
Report process terminates so that while communication efficiency is improved, also may be such that link up between different language user it is more natural and tripping.
Further, since voice signal and/or translation result based on round before current round, to identification text data into
Row translation, the relevance of context during so as to make full use of speech, improves translation accuracy rate.
Finally, due to which languages type corresponding target language type in source can be determined according to default correspondence, can such as be based on
Vocal print feature in targeted voice signal, determines target language type, so as to meet preference and demand of the user in translation,
Realize personalized customization.
An embodiment of the present invention provides a kind of voiced translation processing equipment.Referring to Fig. 5, which includes：Processor
(processor) 501, memory (memory) 502 and bus 503；
Wherein, processor 501 and memory 502 complete mutual communication by bus 503 respectively；
Processor 501 is used to call the programmed instruction in memory 502, and the voice provided with performing above-described embodiment turns over
Translate processing method, such as including：During being reported to the first synthetic speech signal, include part if receiving
The mixing voice signal of one synthetic speech signal, then stop reporting the first synthetic speech signal, the first synthetic speech signal is warp
By obtained after last round of translation and phonetic synthesis；Part first, which is filtered out, from mixing voice signal synthesizes voice letter
Number, current round voice signal to be translated is obtained, and be used as targeted voice signal；Based on targeted voice signal, second is obtained
Synthetic speech signal, and the second synthetic speech signal is reported, the second synthetic speech signal is that targeted voice signal is translated
It is and obtained after phonetic synthesis.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium storing program for executing, the non-transient computer readable storage medium
Matter stores computer instruction, which makes computer perform the voiced translation processing method that above-described embodiment is provided,
Such as including：During being reported to the first synthetic speech signal, voice is synthesized if receiving and including part first
The mixing voice signal of signal, then stop reporting the first synthetic speech signal, the first synthetic speech signal is via last round of time
It is obtained after translation and phonetic synthesis；The first synthetic speech signal of part is filtered out from mixing voice signal, is obtained current
Round voice signal to be translated, and it is used as targeted voice signal；Based on targeted voice signal, the second synthesis voice letter is obtained
Number, and report the second synthetic speech signal, the second synthetic speech signal be targeted voice signal is translated and voice close
Into rear obtained.
One of ordinary skill in the art will appreciate that：Realizing all or part of step of above method embodiment can pass through
The relevant hardware of programmed instruction is completed, and foregoing program can be stored in a computer read/write memory medium, the program
Upon execution, the step of execution includes above method embodiment；And foregoing storage medium includes：ROM, RAM, magnetic disc or light
Disk etc. is various can be with the medium of store program codes.
The embodiments such as information interaction device described above are only schematical, wherein illustrate as separating component
Unit may or may not be physically separate, and thing is may or may not be as the component that unit is shown
Manage unit, you can with positioned at a place, or can also be distributed in multiple network unit.It can select according to the actual needs
Some or all of module therein is selected to realize the purpose of this embodiment scheme.Those of ordinary skill in the art are not paying wound
In the case of the work for the property made, you can to understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
Realized by the mode of software plus required general hardware platform, naturally it is also possible to pass through hardware.Based on such understanding, on
The part that technical solution substantially in other words contributes to the prior art is stated to embody in the form of software product, should
Computer software product can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD, including some fingers
Order is used so that a computer equipment (can be personal computer, server, or network equipment etc.) performs each implementation
Some Part Methods of example or embodiment.
Finally, the present processes are only preferable embodiment, are not intended to limit the protection model of the embodiment of the present invention
Enclose.With within principle, any modification, equivalent replacement, improvement and so on, should be included in all spirit in the embodiment of the present invention
Within the protection domain of the embodiment of the present invention.
Claims (10)
- A kind of 1. voiced translation processing method, it is characterised in that including：During being reported to the first synthetic speech signal, include part the first synthesis voice letter if receiving Number mixing voice signal, then stop reporting first synthetic speech signal, first synthetic speech signal is via upper It is obtained after the translation of one round and phonetic synthesis；Part first synthetic speech signal is filtered out from the mixing voice signal, obtains current round voice to be translated Signal, and it is used as targeted voice signal；Based on the targeted voice signal, the second synthetic speech signal is obtained, and reports second synthetic speech signal, it is described Second synthetic speech signal be the targeted voice signal is translated and phonetic synthesis after it is obtained.
- 2. according to the method described in claim 1, it is characterized in that, first synthetic speech signal and the target language message From the targeted voice signal it is different languages types number for identical languages type or first synthetic speech signal.
- 3. according to the method described in claim 1, it is characterized in that, before report second synthetic speech signal, go back Including：The identification text data to being obtained after targeted voice signal progress speech recognition is obtained, is obtained to the identification text The target text data that data obtain after being translated, and phonetic synthesis is carried out to the target text data, obtain described the Two synthetic speech signals.
- 4. according to the method described in claim 3, it is characterized in that, it is described obtain to it is described identification text data translate after Obtained target text data, including：Determine the corresponding source languages type of the identification text data, and the source languages type is determined according to default correspondence Corresponding target language type；The target language type and the identification text data are inputted to translation coding/decoding model, export the target text Data.
- 5. according to the method described in claim 4, it is characterized in that, described determine the source languages class according to default correspondence The corresponding target language type of type, including：Based on the vocal print feature in the targeted voice signal, the corresponding default languages type of the vocal print feature is determined, to make For the corresponding target language type of the source languages type.
- 6. according to the method described in claim 3, it is characterized in that, it is described obtain to it is described identification text data translate after Obtained target text data, including：If judgement knows that the targeted voice signal and the information that first synthetic speech signal transmits respectively are interrelated, Based on the voice signal and/or translation result of round before current round, the identification text data is translated, to obtain The target text data.
- 7. according to the method described in claim 3, it is characterized in that, it is described to the target text data carry out phonetic synthesis, Second synthetic speech signal is obtained, including：Voice broadcast parameter is obtained, the target text data and the voice broadcast parameter are inputted to phonetic synthesis model, Second synthetic speech signal is exported, when report second synthetic speech signal is included at least in the voice broadcast parameter Used tamber parameter.
- A kind of 8. voiced translation processing unit, it is characterised in that including：Stop broadcasting module, for during being reported to the first synthetic speech signal, including part if receiving The mixing voice signal of first synthetic speech signal, then stop reporting first synthetic speech signal, described first closes It is via obtained after last round of translation and phonetic synthesis into voice signal；Filtering module, for filtering out part first synthetic speech signal from the mixing voice signal, obtains working as front-wheel Secondary voice signal to be translated, and it is used as targeted voice signal；Broadcasting module, for based on the targeted voice signal, obtaining the second synthetic speech signal, and report second synthesis Voice signal, second synthetic speech signal be the targeted voice signal is translated and phonetic synthesis after obtained by 's.
- A kind of 9. voiced translation processing equipment, it is characterised in that including：At least one processor；AndAt least one processor being connected with the processor communication, wherein：The memory storage has the programmed instruction that can be performed by the processor, and the processor calls described program instruction energy Enough perform the method as described in claim 1 to 7 is any.
- 10. a kind of non-transient computer readable storage medium storing program for executing, it is characterised in that the non-transient computer readable storage medium storing program for executing is deposited Computer instruction is stored up, the computer instruction makes the computer perform the method as described in claim 1 to 7 is any.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711107221.9A CN107910004A (en) | 2017-11-10 | 2017-11-10 | Voiced translation processing method and processing device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711107221.9A CN107910004A (en) | 2017-11-10 | 2017-11-10 | Voiced translation processing method and processing device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107910004A true CN107910004A (en) | 2018-04-13 |
Family
ID=61844975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711107221.9A Pending CN107910004A (en) | 2017-11-10 | 2017-11-10 | Voiced translation processing method and processing device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107910004A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109344411A (en) * | 2018-09-19 | 2019-02-15 | 深圳市合言信息科技有限公司 | A kind of interpretation method for listening to formula simultaneous interpretation automatically |
| CN110970014A (en) * | 2019-10-31 | 2020-04-07 | 阿里巴巴集团控股有限公司 | Voice conversion, file generation, broadcast, voice processing method, device and medium |
| CN112652311A (en) * | 2020-12-01 | 2021-04-13 | 北京百度网讯科技有限公司 | Chinese and English mixed speech recognition method and device, electronic equipment and storage medium |
| JP2021513119A (en) * | 2018-04-16 | 2021-05-20 | グーグル エルエルシーＧｏｏｇｌｅ ＬＬＣ | Automated assistants dealing with multiple age groups and / or vocabulary levels |
| WO2021109000A1 (en) * | 2019-12-03 | 2021-06-10 | 深圳市欢太科技有限公司 | Data processing method and apparatus, electronic device, and storage medium |
| WO2021208531A1 (en) * | 2020-04-16 | 2021-10-21 | 北京搜狗科技发展有限公司 | Speech processing method and apparatus, and electronic device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101154220A (en) * | 2006-09-25 | 2008-04-02 | 株式会社东芝 | Machine translation apparatus and method |
| CN103246643A (en) * | 2012-02-10 | 2013-08-14 | 株式会社东芝 | Speech translation apparatus and speech translation method |
| CN105512113A (en) * | 2015-12-04 | 2016-04-20 | 青岛冠一科技有限公司 | Communication type voice translation system and translation method |
| CN106156009A (en) * | 2015-04-13 | 2016-11-23 | 中兴通讯股份有限公司 | Voice translation method and device |
| CN106486125A (en) * | 2016-09-29 | 2017-03-08 | 安徽声讯信息技术有限公司 | A kind of simultaneous interpretation system based on speech recognition technology |
-
2017
- 2017-11-10 CN CN201711107221.9A patent/CN107910004A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101154220A (en) * | 2006-09-25 | 2008-04-02 | 株式会社东芝 | Machine translation apparatus and method |
| CN103246643A (en) * | 2012-02-10 | 2013-08-14 | 株式会社东芝 | Speech translation apparatus and speech translation method |
| CN106156009A (en) * | 2015-04-13 | 2016-11-23 | 中兴通讯股份有限公司 | Voice translation method and device |
| CN105512113A (en) * | 2015-12-04 | 2016-04-20 | 青岛冠一科技有限公司 | Communication type voice translation system and translation method |
| CN106486125A (en) * | 2016-09-29 | 2017-03-08 | 安徽声讯信息技术有限公司 | A kind of simultaneous interpretation system based on speech recognition technology |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2022103191A (en) * | 2018-04-16 | 2022-07-07 | グーグル エルエルシー | Automated assistants that accommodate multiple age groups and/or vocabulary levels |
| US11495217B2 (en) | 2018-04-16 | 2022-11-08 | Google Llc | Automated assistants that accommodate multiple age groups and/or vocabulary levels |
| JP7486540B2 (en) | 2018-04-16 | 2024-05-17 | グーグル エルエルシー | Automated assistants that address multiple age groups and/or vocabulary levels |
| JP2021513119A (en) * | 2018-04-16 | 2021-05-20 | グーグル エルエルシーＧｏｏｇｌｅ ＬＬＣ | Automated assistants dealing with multiple age groups and / or vocabulary levels |
| US11756537B2 (en) | 2018-04-16 | 2023-09-12 | Google Llc | Automated assistants that accommodate multiple age groups and/or vocabulary levels |
| JP7064018B2 (en) | 2018-04-16 | 2022-05-09 | グーグル エルエルシー | Automated assistant dealing with multiple age groups and / or vocabulary levels |
| CN109344411A (en) * | 2018-09-19 | 2019-02-15 | 深圳市合言信息科技有限公司 | A kind of interpretation method for listening to formula simultaneous interpretation automatically |
| CN110970014A (en) * | 2019-10-31 | 2020-04-07 | 阿里巴巴集团控股有限公司 | Voice conversion, file generation, broadcast, voice processing method, device and medium |
| CN110970014B (en) * | 2019-10-31 | 2023-12-15 | 阿里巴巴集团控股有限公司 | Voice conversion, file generation, broadcasting and voice processing method, equipment and medium |
| WO2021109000A1 (en) * | 2019-12-03 | 2021-06-10 | 深圳市欢太科技有限公司 | Data processing method and apparatus, electronic device, and storage medium |
| WO2021208531A1 (en) * | 2020-04-16 | 2021-10-21 | 北京搜狗科技发展有限公司 | Speech processing method and apparatus, and electronic device |
| CN113539233A (en) * | 2020-04-16 | 2021-10-22 | 北京搜狗科技发展有限公司 | Voice processing method and device and electronic equipment |
| CN112652311B (en) * | 2020-12-01 | 2021-09-03 | 北京百度网讯科技有限公司 | Chinese and English mixed speech recognition method and device, electronic equipment and storage medium |
| US11893977B2 (en) | 2020-12-01 | 2024-02-06 | Beijing Baidu Netcom Science Technology Co., Ltd. | Method for recognizing Chinese-English mixed speech, electronic device, and storage medium |
| CN112652311A (en) * | 2020-12-01 | 2021-04-13 | 北京百度网讯科技有限公司 | Chinese and English mixed speech recognition method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107910004A (en) | Voiced translation processing method and processing device | |
| CN111128126B (en) | Multi-language intelligent voice conversation method and system | |
| CN105096941B (en) | Audio recognition method and device | |
| US8914294B2 (en) | System and method of providing an automated data-collection in spoken dialog systems | |
| CN102723078B (en) | Emotion speech recognition method based on natural language comprehension | |
| Bell et al. | Prosodic adaptation in human-computer interaction | |
| Hjalmarsson | The additive effect of turn-taking cues in human and synthetic voice | |
| CN107680597A (en) | Audio recognition method, device, equipment and computer-readable recording medium | |
| CN103903627A (en) | Voice-data transmission method and device | |
| CN108831439A (en) | Voice recognition method, device, equipment and system | |
| CN110970018B (en) | Speech recognition method and device | |
| CN102982811A (en) | Voice endpoint detection method based on real-time decoding | |
| CN103003876A (en) | Modification of speech quality in conversations over voice channels | |
| CN102903361A (en) | Instant call translation system and instant call translation method | |
| CN103065620A (en) | Method with which text input by user is received on mobile phone or webpage and synthetized to personalized voice in real time | |
| CN108986798B (en) | Processing method, device and the equipment of voice data | |
| CN104766608A (en) | Voice control method and voice control device | |
| CN106875936A (en) | Voice recognition method and device | |
| CN112131359A (en) | Intention identification method based on graphical arrangement intelligent strategy and electronic equipment | |
| CN1714390B (en) | Speech recognition device and method | |
| CN114818649A (en) | Service consultation processing method and device based on intelligent voice interaction technology | |
| CN107886940A (en) | Voiced translation processing method and processing device | |
| CN102056093A (en) | Method for converting text message into voice message | |
| CN108364655A (en) | Method of speech processing, medium, device and computing device | |
| CN112216270B (en) | Speech phoneme recognition method and system, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180413 |
|
| RJ01 | Rejection of invention patent application after publication |