CN106057196B - Vehicle voice data parses recognition methods - Google Patents
Vehicle voice data parses recognition methods Download PDFInfo
- Publication number
- CN106057196B CN106057196B CN201610534783.0A CN201610534783A CN106057196B CN 106057196 B CN106057196 B CN 106057196B CN 201610534783 A CN201610534783 A CN 201610534783A CN 106057196 B CN106057196 B CN 106057196B
- Authority
- CN
- China
- Prior art keywords
- model
- state
- parameter
- vector
- sequence
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 4
- 239000000203 mixture Substances 0.000 claims description 8
- 238000005520 cutting process Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 4
- 150000001875 compounds Chemical class 0.000 claims description 2
- 241001269238 Data Species 0.000 abstract description 5
- 238000012360 testing method Methods 0.000 description 10
- 230000011218 segmentation Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000004378 air conditioning Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000000572 ellipsometry Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 229940068517 fruit extracts Drugs 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation 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
- G10L15/08—Speech classification or search
- G10L15/14—Speech classification or search using statistical models, e.g. Hidden Markov Models [HMMs]
- G10L15/142—Hidden Markov Models [HMMs]
-
- 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
- G10L15/08—Speech classification or search
- G10L15/14—Speech classification or search using statistical models, e.g. Hidden Markov Models [HMMs]
- G10L15/142—Hidden Markov Models [HMMs]
- G10L15/144—Training of HMMs
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Machine Translation (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention provides a kind of vehicle voice datas to parse recognition methods, this method comprises: reading in the voice to be identified, observation sequence is obtained after front-end processing, the conditional probability of calculating observation sequence model corresponding with all entries determines entry recognition result according to the conditional probability.The invention proposes a kind of vehicle voice datas to parse recognition methods, does not need the label training sample set in offline dictionary, small to the dependence of rule, improves accuracy of identification, adapts to the demand that onboard system is constantly updated.
Description
Technical field
The present invention relates to speech recognition, in particular to a kind of vehicle voice data parses recognition methods.
Background technique
The development that the technologies such as cloud computing, big data, data mining will push information services industry faster and better, wherein merging
Guidance people are more accurately and efficiently obtained required information and service by the information service of natural language understanding technology.Voice
Man-machine interaction mode of the technology as most desirable will gradually become one kind more and more crucial in many interactive modes.
Specific to automotive field, the practical intelligent information service system of height is customized out using natural language understanding technology, with more people
Property man-machine interaction mode more convenient, accurate information speech order and navigation are provided, will be prospect for the experience of driving
Wide promotion.However existing vehicle environment speech recognition is by the way that needing in biggish offline dictionary, largely label trains sample
This collection carries out study and carries out semantic supposition, relatively high to the dependence of rule, less flexibly, it is continuous can not to adapt to onboard system
The demand and precision of variation and accuracy are lower.
Summary of the invention
To solve the problems of above-mentioned prior art, the invention proposes a kind of vehicle voice datas to parse identification side
Method, comprising:
The voice to be identified is read in, observation sequence, calculating observation sequence and all entries pair are obtained after front-end processing
The conditional probability for answering model determines entry recognition result according to the conditional probability.
Preferably, before the conditional probability of calculating observation sequence model corresponding with all entries, further includes:
Speech primitive corresponding to characteristic parameter sequence to be identified is estimated, the primitive includes word, syllable, initial consonant, rhythm
Characteristic parameter sequence is converted to Recognition unit by mother；The process for establishing model includes:
(1), initial parameter value is randomly choosed, HMM model λ is initialized；
It (2), is that each state carries out cutting by observation sequence；The result of cutting is exactly the corresponding observation of each state
Frame set；
(3), the observation vector set for belonging to each state is divided into M cluster using Segment Clustering algorithm, M is Gaussian Mixture
Degree, a single Gaussian parameter of every cluster for Gaussian-mixture probability density, carries out the estimation of following parameter later:
cjk=the vector number that kth set is in when state j/belongs to the vector sum of state j；
μjkThe vector sample average of kth set is in under=state j；
UjkThe vector sample covariance matrix of kth set is in under=state j；
The HMM model λ ' of a update is obtained according to above-mentioned parameter；
(4), model λ ' is compared with initial model λ, if model difference is more than preset threshold value, by mould
Type λ replaces with λ ' and repeats the above steps 2 and 3；If difference is lower than threshold value, it is determined that restrained for model, saving should
Model；
By above-mentioned iteration so as to constantly be corrected during entire model training to initial parameter value.
The present invention compared with prior art, has the advantage that
The invention proposes a kind of vehicle voice datas to parse recognition methods, does not need the label training sample in offline dictionary
This collection, it is small to the dependence of rule, accuracy of identification is improved, the demand that onboard system is constantly updated is adapted to.
Detailed description of the invention
Fig. 1 is the flow chart of vehicle voice data parsing recognition methods according to an embodiment of the present invention.
Specific embodiment
Retouching in detail to one or more embodiment of the invention is hereafter provided together with the attached drawing of the diagram principle of the invention
It states.The present invention is described in conjunction with such embodiment, but the present invention is not limited to any embodiments.The scope of the present invention is only by right
Claim limits, and the present invention covers many substitutions, modification and equivalent.Illustrate in the following description many details with
Just it provides a thorough understanding of the present invention.These details are provided for exemplary purposes, and without in these details
Some or all details can also realize the present invention according to claims.
An aspect of of the present present invention provides a kind of vehicle voice data parsing recognition methods.Fig. 1 is to implement according to the present invention
The vehicle voice data of example parses recognition methods flow chart.
Onboard system of the invention is composed of identification module, semantic supposition categorization module.Utilize machine learning method
Effective study is carried out to establish segmenter to training corpus.Word mark is carried out using CRF.Then semantic supposition is carried out.
Classification to proper noun, storage and tissue and voice command convenient for proper noun.
The identification process of identification module can be described as: read in the voice to be identified, after front-end processing, the sight that will obtain
It surveys sequence X to match all entries, that is, design conditions probability, wherein the corresponding entry of the model of maximum probability is just
It is recognition result.Complete above-mentioned identification, it is necessary to first complete learning model training.
Go out speech primitive corresponding to characteristic parameter sequence to be identified from probability angle estimation, the primitive include word, syllable,
Initial consonant, simple or compound vowel of a Chinese syllable, so that characteristic parameter sequence is converted to Recognition unit.The process for establishing model includes:
1, initial parameter value is randomly choosed, HMM model λ is initialized.
It 2, is that each state carries out cutting by observation sequence.The result of cutting is exactly that each state corresponds to an observation frame
Set.
3, the observation vector set for belonging to each state is divided into M cluster using segmentation K mean algorithm, M is Gaussian Mixture
Degree, a single Gaussian parameter of every cluster for Gaussian-mixture probability density, carries out the estimation of following parameter later.
cjk=the vector number that kth set is in when state j/belongs to the vector sum of state j
μjkThe vector sample average of kth set is in under=state j
UjkThe vector sample covariance matrix of kth set is in under=state j
The HMM model λ ' of a update is obtained according to above-mentioned parameter；
4, model λ ' is compared with initial model λ, if model difference is more than preset threshold value, by model
λ replaces with λ ' and repeats the above steps 2 and 3；If difference is lower than threshold value, it is determined that restrained for model, saved the mould
Type.
By above-mentioned iteration so as to constantly be corrected during entire model training to initial parameter value.
In the training stage of model, directly by MFCC characteristic parameter as observation, a MFCC vector is exactly a sight
Measured value.Secondly the parameter b of composition probability density function is calculatedj(o):
bj(o)=∑ cjmM(o,μjm,Ujm), 1 < j < N
Wherein o indicates observation, cjmM-th of mixed coefficint of state j is represented, M is ellipsometry density function, by putting down
Mean vector μjmWith covariance matrix Ujm；
Model training process is as follows:
(1), all training characteristics parameters are sliced into each state.
(2), the single Gauss model that the characteristic parameter that each state is possessed is given to mixed Gauss model, to model into
The following amendment of row:
Wherein γt(j, k) isL is sample size；CtFor t moment scale factor.
(3), judge whether to restrain.Terminate if convergence, if not satisfied, continuing iteration.
In classifying to proper noun.The training sample set and test set of database are obtained first.To training sample
This collection be trained on the basis of pretreatment and text representation to training sample set, obtains a classifier.It is assessing
The classifier stage tests test set.It after being pre-processed to training sample, is segmented, by each proper noun
It is converted to the vector being made of morpheme.The word frequency and reverse frequency of each morpheme are counted using training sample, and are thus calculated every
Regularization word frequency and reverse frequency ratio of a morpheme to classification predetermined, as word d to the power of corresponding classification i
Value wi(d)。
Wherein N is proper noun sum, niFor the proper noun quantity comprising entry i；During the test, it counts respectively
Proper noun to be processed provides final classification results to the weights sum of each classification.
Wherein ZY is proper noun to be sorted, and M is classification number ZyjJ-th of morpheme in as ZY.
Four contents are arranged in every voice command in database, are intention, voice command original text, voice command respectively
Segment information, the word mark information of voice command.Generate the training file of participle, the test file of participle, word mark
The test file of training file, word mark.The present invention segments first to voice command and word mark.For repeatability
Mistake, it may be found that mistake be added to modifying for batch in the customized dictionary of program.
Before participle, the problem of participle, is converted into sequence mark problem.The beginnings of tagged words, the centre of word, word
The word that ending and single word are constituted, the feature for then needing to learn using default template definition, every a line generation in template file
One template of table, what the row in macro [row, col] in each template was represented is opposite line number, and what col was indicated is absolute
Line number.The template according to defined in template file, training file is according to these template generation Feature Words.
Using the learning training for segmenting training file progress condition random field, a Words partition system is obtained.Utilize part of speech mark
Remember that training file carries out the learning training of condition random field, obtains a word mark system.It is surveyed with participle test file
Examination, obtains the precision of participle.Word mark test file is tested, the precision of word mark system is obtained.
After participle terminates, the result that will be segmented is needed to carry out transformation appropriate in order to later use word mark system
System carries out word mark.For every voice command, Words partition system exports B1 word segmentation result.The part of speech mark obtained using training
Note system carries out word mark to file, and every voice command can obtain B1 word segmentation result, this step is for each input
There is the output of B2 word mark as a result, therefore every voice command finally can get B1*B2 recognition result, is screened out from it most
B excellent recognition result, the participle of database Plays and word mark information are write back.It is same with word mark result when segmenting
When it is completely the same when, determine that recognition result is correct.
It further comprise respectively tying each participle wherein in generating B1 word segmentation result and B2 word mark result
The probability of fruit extracts, and is stored in array p1In, the probability of word mark result, which extracts, is stored in p2In array.For
B1*B2 recognition result of every voice command can pass through p1、p2The two arrays calculate its generating probability, calculating process are as follows:
P [i]=p2[i]*p1[i/B2] i=0,1,2 ..., B1*B2-1
The generating probability p [i] of this B1*B2 recognition result is ranked up, the highest B recognition result of output probability.
After completing training study, output module exports the participle and word mark result of given voice command.For
The voice command of input searches dictionary, searches special noun in voice command, corresponding special symbol is replaced with if finding
Number, and file 1 is written；The proper noun found is identified with bracket at the same time, and file 2 is written.File 1 is converted
The input format of test module is segmented for condition random field and file 3 is written.File 3 is segmented, the result of participle is protected
It is stored in file 4.The input format of word mark system is converted by the result of participle and file 5 is written.For every voice
The B1 word segmentation result generated is ordered, the probability of word segmentation result is saved in p1.Part of speech is carried out to the voice command in file 5
Label.As a result it is saved in file 6.B1*B2 recognition result is obtained for every voice command altogether at this time, needs to export most
B good recognition result.B final result is written into file 7.Convert file 7 to finally specified output format.
Semantic understanding uses the method based on statistical learning in the present invention.Semantic major class in onboard system includes navigation road
Line, road conditions are answered and are made a phone call, air-conditioning adjusting, the functions such as weather voice command, radio.Part of semanteme also needs
It takes parameter: such as making a phone call needs and know the specific telephone number dialed.Semantic supposition problem of the invention can be converted into
The intention of the text of input is assigned in predefined intention class.The type of voice command intention is speculated first, if intention needs
Further parameter then finds corresponding parameter in voice command.
To the following conditions probabilistic Modeling in word mark problem:
p(s1…sm|x1…xm), for x1…xmRepresent the single word in the voice command of input, and s1…sm∈ S generation
The all possible part of speech combination of table.For one by x1…xmThe voice command of composition shares kmKind word mark combination.K=| S
|.Therefore this k is establishedmThe probability distribution of kind word mark result.
Then it obtains:
The finite aggregate feature vector for indicating predefined set of words X Yu tag set Y, passes through s1…smAnd x1…xm's
A large amount of training, obtain parameter vector w, finally obtain p (s1…sm|x1…xm)。
It completes to seek for inputting x after training1…xmState s1…sm, that is, it solves:
The word after being characterized in voice command participle is used to be also possible to the combination of corresponding part of speech or the two.
After choosing series of features, feature is added according to training sample, adjusts weight.
In the text classification problem for semantics recognition, for giving the training sample of set of words X and tag set Y
(xi,yi), i=l ..., n, n are total sample number, following optimization problem is set:
When test, for point x according to wTX > 0 provides classification results.
In conclusion the invention proposes a kind of vehicle voice datas to parse recognition methods, do not need in offline dictionary
Training sample set is marked, it is small to the dependence of rule, accuracy of identification is improved, the demand that onboard system is constantly updated is adapted to.
Obviously, it should be appreciated by those skilled in the art, each module of the above invention or each steps can be with general
Computing system realize that they can be concentrated in single computing system, or be distributed in multiple computing systems and formed
Network on, optionally, they can be realized with the program code that computing system can be performed, it is thus possible to they are stored
It is executed within the storage system by computing system.In this way, the present invention is not limited to any specific hardware and softwares to combine.
It should be understood that above-mentioned specific embodiment of the invention is used only for exemplary illustration or explains of the invention
Principle, but not to limit the present invention.Therefore, that is done without departing from the spirit and scope of the present invention is any
Modification, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.In addition, appended claims purport of the present invention
Covering the whole variations fallen into attached claim scope and boundary or this range and the equivalent form on boundary and is repairing
Change example.
Claims (1)
1. a kind of vehicle voice data parses recognition methods characterized by comprising
The voice to be identified is read in, observation sequence, calculating observation sequence mould corresponding with all entries are obtained after front-end processing
The conditional probability of type determines entry recognition result according to the conditional probability；
Before the conditional probability of the calculating observation sequence model corresponding with all entries, further includes:
Speech primitive corresponding to characteristic parameter sequence to be identified is estimated, the primitive includes word, syllable, initial consonant, simple or compound vowel of a Chinese syllable, is incited somebody to action
Characteristic parameter sequence is converted to Recognition unit；The process for establishing model includes:
(1), initial parameter value is randomly choosed, HMM model λ is initialized；
It (2), is that each state carries out cutting by observation sequence；The result of cutting is exactly that each state corresponds to an observation frame collection
It closes；
(3), the observation vector set for belonging to each state being divided into M cluster using Segment Clustering algorithm, M is Gaussian Mixture degree,
A single Gaussian parameter of every cluster for Gaussian-mixture probability density, carries out the estimation of following parameter later:
cjk=the vector number that kth set is in when state j/belongs to the vector sum of state j；
μjkThe vector sample average of kth set is in under=state j；
UjkThe vector sample covariance matrix of kth set is in under=state j；
The HMM model λ ' of a update is obtained according to above-mentioned parameter；
(4), model λ ' is compared with initial model λ, if model difference is more than preset threshold value, by model λ
It replaces with λ ' and repeats the above steps 2 and 3；If difference is lower than threshold value, it is determined that restrained for model, saved the mould
Type；
By above-mentioned iteration so as to constantly be corrected during entire model training to initial parameter value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610534783.0A CN106057196B (en) | 2016-07-08 | 2016-07-08 | Vehicle voice data parses recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610534783.0A CN106057196B (en) | 2016-07-08 | 2016-07-08 | Vehicle voice data parses recognition methods |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106057196A CN106057196A (en) | 2016-10-26 |
| CN106057196B true CN106057196B (en) | 2019-06-11 |
Family
ID=57184974
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610534783.0A Expired - Fee Related CN106057196B (en) | 2016-07-08 | 2016-07-08 | Vehicle voice data parses recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106057196B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971721A (en) * | 2017-03-29 | 2017-07-21 | 沃航（武汉）科技有限公司 | A kind of accent speech recognition system based on embedded mobile device |
| CN108986811B (en) * | 2018-08-31 | 2021-05-28 | 北京新能源汽车股份有限公司 | Voice recognition detection method, device and equipment |
| CN111353292B (en) * | 2020-02-26 | 2023-06-16 | 支付宝(杭州)信息技术有限公司 | Analysis method and device for user operation instruction |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1490786A (en) * | 2002-10-17 | 2004-04-21 | 中国科学院声学研究所 | Phonetic recognition confidence evaluating method, system and dictation device therewith |
| CN101930735A (en) * | 2009-06-23 | 2010-12-29 | 富士通株式会社 | Speech emotion recognition equipment and speech emotion recognition method |
| CN101980336B (en) * | 2010-10-18 | 2012-01-11 | 福州星网视易信息系统有限公司 | Hidden Markov model-based vehicle sound identification method |
| CN103065626A (en) * | 2012-12-20 | 2013-04-24 | 中国科学院声学研究所 | Automatic grading method and automatic grading equipment for read questions in test of spoken English |
| CN103810998A (en) * | 2013-12-05 | 2014-05-21 | 中国农业大学 | Method for off-line speech recognition based on mobile terminal device and achieving method |
| CN105390133A (en) * | 2015-10-09 | 2016-03-09 | 西北师范大学 | Tibetan TTVS system realization method |
-
2016
- 2016-07-08 CN CN201610534783.0A patent/CN106057196B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1490786A (en) * | 2002-10-17 | 2004-04-21 | 中国科学院声学研究所 | Phonetic recognition confidence evaluating method, system and dictation device therewith |
| CN101930735A (en) * | 2009-06-23 | 2010-12-29 | 富士通株式会社 | Speech emotion recognition equipment and speech emotion recognition method |
| CN101980336B (en) * | 2010-10-18 | 2012-01-11 | 福州星网视易信息系统有限公司 | Hidden Markov model-based vehicle sound identification method |
| CN103065626A (en) * | 2012-12-20 | 2013-04-24 | 中国科学院声学研究所 | Automatic grading method and automatic grading equipment for read questions in test of spoken English |
| CN103810998A (en) * | 2013-12-05 | 2014-05-21 | 中国农业大学 | Method for off-line speech recognition based on mobile terminal device and achieving method |
| CN105390133A (en) * | 2015-10-09 | 2016-03-09 | 西北师范大学 | Tibetan TTVS system realization method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106057196A (en) | 2016-10-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106407333B (en) | Spoken language query identification method and device based on artificial intelligence | |
| CN110210029A (en) | Speech text error correction method, system, equipment and medium based on vertical field | |
| CN108763510A (en) | Intension recognizing method, device, equipment and storage medium | |
| CN108108351A (en) | A kind of text sentiment classification method based on deep learning built-up pattern | |
| CN105205124B (en) | A kind of semi-supervised text sentiment classification method based on random character subspace | |
| CN107330011A (en) | The recognition methods of the name entity of many strategy fusions and device | |
| CN111046670B (en) | Entity and relationship combined extraction method based on drug case legal documents | |
| CN104199965A (en) | Semantic information retrieval method | |
| CN106294344A (en) | Video retrieval method and device | |
| CN106340297A (en) | Speech recognition method and system based on cloud computing and confidence calculation | |
| CN103678271B (en) | A kind of text correction method and subscriber equipment | |
| CN112016313B (en) | Spoken language element recognition method and device and warning analysis system | |
| CN110992988B (en) | Speech emotion recognition method and device based on domain confrontation | |
| CN113408287B (en) | Entity identification method and device, electronic equipment and storage medium | |
| CN106057196B (en) | Vehicle voice data parses recognition methods | |
| CN106528776A (en) | Text classification method and device | |
| CN110046264A (en) | A kind of automatic classification method towards mobile phone document | |
| CN113919366A (en) | Semantic matching method and device for power transformer knowledge question answering | |
| CN106202045B (en) | Special audio recognition method based on car networking | |
| CN114154570A (en) | Sample screening method and system and neural network model training method | |
| CN108681532B (en) | Sentiment analysis method for Chinese microblog | |
| CN110097096A (en) | A kind of file classification method based on TF-IDF matrix and capsule network | |
| CN112417132A (en) | New intention recognition method for screening negative samples by utilizing predicate guest information | |
| CN106203520B (en) | SAR image classification method based on depth Method Using Relevance Vector Machine | |
| CN112489689B (en) | Cross-database voice emotion recognition method and device based on multi-scale difference countermeasure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190611 Termination date: 20210708 |