CN110265034A - A kind of power grid regulation auto-answer method - Google Patents
A kind of power grid regulation auto-answer method Download PDFInfo
- Publication number
- CN110265034A CN110265034A CN201910295877.0A CN201910295877A CN110265034A CN 110265034 A CN110265034 A CN 110265034A CN 201910295877 A CN201910295877 A CN 201910295877A CN 110265034 A CN110265034 A CN 110265034A
- Authority
- CN
- China
- Prior art keywords
- power grid
- grid regulation
- language
- answer method
- identification model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 230000004044 response Effects 0.000 claims abstract description 12
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 9
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 9
- 230000001755 vocal effect Effects 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 239000000463 material Substances 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 6
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 238000004891 communication Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 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
- G10L13/00—Speech synthesis; Text to speech systems
- G10L13/08—Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
- G10L13/10—Prosody rules derived from text; Stress or intonation
-
- 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
- G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
- G10L15/063—Training
-
- 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
- G10L15/16—Speech classification or search using artificial neural networks
-
- 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
- G10L17/00—Speaker identification or verification techniques
-
- 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
- G10L17/00—Speaker identification or verification techniques
- G10L17/04—Training, enrolment or model building
-
- 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
- G10L17/00—Speaker identification or verification techniques
- G10L17/18—Artificial neural networks; Connectionist approaches
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M2250/00—Details of telephonic subscriber devices
- H04M2250/74—Details of telephonic subscriber devices with voice recognition means
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Telephonic Communication Services (AREA)
Abstract
The invention discloses a kind of power grid regulation auto-answer methods, comprising the following steps: S01: establishes identification model；S02: identifying call, meets automatic-answering back device condition and then carries out step S03, on the contrary then carry out step S04；S03: according to language content, carrying out response using synthesis voice, and end of conversation then gos to step S05, otherwise return step S02；S04: artificial treatment is gone to；S05: recording call content is simultaneously input to identification model.Phone is identified by identification model, supports to be forwarded to artificial treatment under specific circumstances, utmostly improves and convey efficiency, while ensure that higher accuracy, improve the working efficiency of power grid regulation.Substantial effect of the invention is: transmission efficiency is high, and identification method is accurate, answering flexibility and reliability, is conducive to the working efficiency for improving power grid regulation.
Description
Technical field
Originally practical to be related to field of communication technology, and in particular to a kind of power grid regulation auto-answer method.
Background technique
According to dispatching of power netwoks regulation, distributed power station needs power station operations staff to pass through electricity when being related to changes of operating modes
It talks about and is applied to dispatching of power netwoks personnel, relevant operation could be executed after reply, and after having executed, it is also necessary to lead to again
Phone is crossed to dispatching of power netwoks personnel reporting operations as a result, dispatching of power netwoks personnel is according to session results, the unit that timely updates runs shape
State and dispatching log.The Primary communication mode of existing power plants and grid coordination is to pass through electricity in advance when mode changes when power station using phone
It talks about and is reported to scheduling, after informed consent, then ways of carrying out change, capable report is transferred in area by phone again after change.Tradition
Use manual type receive calls, then carry out hand-kept, not only increase dispatcher's workload, but also the record side of papery
Formula is unfavorable for the propagation of data, is unable to give full play the value of data, and the data of preservation can only be utilized in a small range, nothing
Method is for quickly analyzing and further further investigation.
A kind of anti-telephone fraud system based on speech recognition of the disclosure of the invention of Authorization Notice No. CN103685613B and
Its method, the system is including starting switch, main control module, voice obtain module, memory module, remote communication module, phone
Swindle keywords database and cue module.Wherein, main control module is used for crucial to speech feature extraction, keyword identification, swindle
Word compares and result judgement；Remote communication module is for remotely updating telephone fraud keywords database and needing to upload in real time
Data transmit-receive function when data；Telephone fraud keywords database swindles several keyword structures extracted in case by common telephone
At.The invention can help telephone subscriber to identify the telephone fraud to emerge one after another.
The identification of the prior art there are accuracys bad, the problem of recognition effect difference, not to the help of power grid regulation efficiency
Greatly.
Summary of the invention
Exist for the prior art and identify that accuracy is bad, the problem of recognition effect difference, the present invention provides a kind of power grids
Regulate and control auto-answer method, the acoustic information in phone can be accurately identified, and artificial treatment can be combined when that can not identify,
Recognition efficiency and accuracy are improved, the working efficiency for improving power grid regulation is helped.
It is technical solution of the present invention below.
A kind of power grid regulation auto-answer method, comprising the following steps: S01: establish identification model；S02: call is carried out
Identification meets automatic-answering back device condition and then carries out step S03, on the contrary then carry out step S04；S03: according to language content, conjunction is used
Response is carried out at voice, and end of conversation then gos to step S05, otherwise return step S02；S04: artificial treatment is gone to；S05:
Recording call content is simultaneously input to identification model.Phone is identified by identification model, supports switching under specific circumstances
It to artificial treatment, utmostly improves and conveys efficiency, while ensure that higher accuracy, improve the work effect of power grid regulation
Rate.
Preferably, the process of the step S01 includes: to establish acoustic model and language model, by neural network side
Formula usage history dialog context trains two kinds of models.It is by the mode of neural network learning, acoustic model and language model is complete
It is kind, and due to using history to converse, there is stronger specific aim, have to the common sentence and vocal print of power grid regulation personnel
There is higher recognition capability.
Preferably, the acoustic model includes the vocal print feature of regulation personnel.Vocal print feature information is directly adjusted with corresponding
Control personnel pairing, specific aim with higher.
Preferably, the detailed process of the step S02 includes: to identify to vocal print feature, and it is eligible, it carries out
Language content identification, if vocal print is ineligible, then go to step S04；As the discrimination of language content lower than threshold value or goes out
Existing language-specific instruction, then go to step S04.Authentication is played the role of in the identification of vocal print first, if can not identify,
Special circumstances are then represented, artificial treatment is needed, in addition if when language content discrimination is poor, illustrate to be likely to occur new
The problem of, it is therefore desirable to artificial treatment, to improve the reliability of work, wherein language-specific instruction includes that " switching is artificial " etc. is bright
Really indicate the language message for needing manually to help.
Preferably, the detailed process of the step S03 includes: to be classified and marked according to key word information in language
Note is assert end of conversation if dropped calls or appearance terminate sound instruction, is jumped according to classification results using synthesis voice answer-back
Step S05 is gone to, otherwise return step S02.Since synthesis voice will appear the problem of response inaccuracy in response, so needing
Retain certain means to terminate synthesis voice answer-back if necessary.
Preferably, the new keywords occurred in manual procedure are marked to identification model in the step S04
In.It is exactly to identify unsuccessful due to being transferred to artificial treatment, the dialog context of artificial treatment is to need to analyze instruction
Experienced emphasis, collecting new keywords is a kind of effective means.
Preferably, the detailed process of the step S05 includes: recording call content, it is right using dialog context as material
Identification model is trained.The content conversed every time is all the most effective material of training.
Substantial effect of the invention is: transmission efficiency is high, and identification method is accurate, and answering flexibility and reliability is conducive to
Improve the working efficiency of power grid regulation.
Specific embodiment
The technical program is further elaborated below in conjunction with specific embodiment.
Embodiment:
A kind of power grid regulation auto-answer method, comprising the following steps:
S01: identification model is established；Acoustic model and language model are established, neural network fashion usage history dialog context is passed through
Two kinds of models of training.It is by the mode of neural network learning, acoustic model and language model is perfect, and due to using
History call, has stronger specific aim, to common sentence and the vocal print recognition capability with higher of power grid regulation personnel.Its
Middle acoustic model includes the vocal print feature of regulation personnel.Vocal print feature information is directly matched with corresponding regulation personnel, is had higher
Specific aim
S02: identifying call, meets automatic-answering back device condition and then carries out step S03, on the contrary then carry out step S04；To vocal print
Feature is identified, eligible, carries out language content identification, and if vocal print is ineligible, then go to step S04；Such as language
Lower than threshold value or there is language-specific instruction in the discrimination of speech content, then go to step S04.Body is played in the identification of vocal print first
The effect of part verifying represents special circumstances if can not identify, needs artificial treatment, in addition if language content identifies
When rate is poor, illustrate to be likely to occur new problem, it is therefore desirable to artificial treatment, to improve the reliability of work, wherein special
Determining sound instruction includes that the language message for needing manually to help is explicitly indicated in " switching is artificial " etc..
S03: according to language content, carrying out response using synthesis voice, and end of conversation then gos to step S05, otherwise returns
Return step S02；Classification and marking is carried out according to key word information in language, according to classification results using synthesis voice answer-back, such as
Dropped calls or appearance terminate sound instruction and then assert end of conversation, and go to step S05, otherwise return step S02.Due to closing
It will appear the problem of response inaccuracy in response at voice, so needing to retain certain means to terminate synthesis if necessary
Voice answer-back.
S04: artificial treatment is gone to；The new keywords occurred in manual procedure are marked into identification model.Due to
The reason of being transferred to artificial treatment is exactly to identify unsuccessful, therefore the dialog context of artificial treatment is the weight for needing analyzing and training
Point, collecting new keywords is a kind of effective means.
S05: recording call content is simultaneously input to identification model.Recording call content, using dialog context as material, to knowledge
Other model is trained.The content conversed every time is all the most effective material of training.
Wherein identification model converts-perceives linear prediction as feature vector using relative spectral, uses Gaussian Mixture mould
Type-hidden Markov model trains these models with maximum-likelihood criterion and expectation-maximization algorithm.
Wherein since regulation call voice response requires response news speed, the meeting in speed by database mode retrieval technique
Response is influenced, for quick response user, Hash Feature Mapping algorithm is used, quickly obtains query result.Feature Hash method
Target be original high dimensional feature vector compression into lower dimensional feature vector, and do not lose the expression energy of primitive character as far as possible
Power.Intention, entity are regarded as a data set, by hash algorithm, obtain corresponding Key, it can be with quick obtaining Key index institute
Corresponding data set.
Before being handled, by initialization, the cryptographic Hash of substation, name and its entity can be calculated, is established
Memory Hash table directly calculates the uncommon value of conjunction, is index with this value after being intended to identify, can be with quick-searching to accordingly
The content of the needs such as power transformation station name, name retrieval.So as to avoid time delay caused by each searching database.
The detailed process for wherein synthesizing voice is, by the text generated after input text or speech recognition, to carry out text
This analysis carries out simple punctuate processing first, then carries out further participle operation according to grammer and language rule.Finally
The differentiation of polyphone is carried out according to semantic analysis, some proprietary vocabulary can be put into dictionary and carry out special and reason.After text analyzing
Then the sound auxiliary sequence of available sentence to be synthesized carries out rhythm processing.According to counted on training set fundamental frequency, when
The priori knowledges such as length, frequency spectrum, the corresponding prosodic information of available text to be synthesized.Analysis is finally obtained into the logical ginseng of prosodic information
Number synthesizer and Cheng Yuyin.
Phone is identified by identification model, supports to be forwarded to artificial treatment under specific circumstances, utmostly mention
Height conveys efficiency, while ensure that higher accuracy, improves the working efficiency of power grid regulation.
It should be noted that the specific embodiment is only used for that technical solution is further described, it is not used in and limits the skill
The range of art scheme, any modifications, equivalent substitutions and improvements etc. based on this technical solution are regarded as in protection of the invention
In range.
Claims (7)
1. a kind of power grid regulation auto-answer method, which comprises the following steps:
S01: identification model is established；
S02: identifying call, meets automatic-answering back device condition and then carries out step S03, on the contrary then carry out step S04；
S03: according to language content, carrying out response using synthesis voice, and end of conversation then gos to step S05, otherwise returns to step
Rapid S02；
S04: artificial treatment is gone to；
S05: recording call content is simultaneously input to identification model.
2. a kind of power grid regulation auto-answer method according to claim 1, which is characterized in that the mistake of the step S01
Journey includes: to establish acoustic model and language model, passes through neural network fashion usage history dialog context two kinds of models of training.
3. a kind of power grid regulation auto-answer method according to claim 2, which is characterized in that the acoustic model includes
The vocal print feature of regulation personnel.
4. a kind of power grid regulation auto-answer method according to claim 2 or 3, which is characterized in that the step S02's
Detailed process includes: to identify to vocal print feature, eligible, carries out language content identification, if vocal print is ineligible,
Then go to step S04；If the discrimination of language content lower than threshold value or language-specific instruction occurs, then go to step S04.
5. a kind of power grid regulation auto-answer method according to claim 1, which is characterized in that the tool of the step S03
Body process includes: to carry out classification and marking according to key word information in language, according to classification results using synthesis voice answer-back, such as
Dropped calls or appearance terminate sound instruction and then assert end of conversation, and go to step S05, otherwise return step S02.
6. a kind of power grid regulation auto-answer method according to claim 1, which is characterized in that, will in the step S04
The new keywords occurred in manual procedure are marked into identification model.
7. a kind of power grid regulation auto-answer method according to claim 1, which is characterized in that the tool of the step S05
Body process includes: recording call content, using dialog context as material, is trained to identification model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910295877.0A CN110265034A (en) | 2019-04-12 | 2019-04-12 | A kind of power grid regulation auto-answer method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910295877.0A CN110265034A (en) | 2019-04-12 | 2019-04-12 | A kind of power grid regulation auto-answer method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110265034A true CN110265034A (en) | 2019-09-20 |
Family
ID=67913556
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910295877.0A Pending CN110265034A (en) | 2019-04-12 | 2019-04-12 | A kind of power grid regulation auto-answer method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110265034A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992941A (en) * | 2019-10-22 | 2020-04-10 | 国网天津静海供电有限公司 | Power grid dispatching voice recognition method and device based on spectrogram |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103811009A (en) * | 2014-03-13 | 2014-05-21 | 华东理工大学 | Smart phone customer service system based on speech analysis |
| CN105895101A (en) * | 2016-06-08 | 2016-08-24 | 国网上海市电力公司 | Speech processing equipment and processing method for power intelligent auxiliary service system |
| CN106328166A (en) * | 2016-08-31 | 2017-01-11 | 上海交通大学 | Man-machine dialogue anomaly detection system and method |
| CN107278302A (en) * | 2017-03-02 | 2017-10-20 | 深圳前海达闼云端智能科技有限公司 | A kind of robot interactive method and interaction robot |
| CN109119084A (en) * | 2018-07-13 | 2019-01-01 | 广东电网有限责任公司 | A kind of dispatch call method and system based on speech recognition |
| CN109413286A (en) * | 2018-10-22 | 2019-03-01 | 北京移数通电讯有限公司 | A kind of intelligent customer service voice response system and method |
-
2019
- 2019-04-12 CN CN201910295877.0A patent/CN110265034A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103811009A (en) * | 2014-03-13 | 2014-05-21 | 华东理工大学 | Smart phone customer service system based on speech analysis |
| CN105895101A (en) * | 2016-06-08 | 2016-08-24 | 国网上海市电力公司 | Speech processing equipment and processing method for power intelligent auxiliary service system |
| CN106328166A (en) * | 2016-08-31 | 2017-01-11 | 上海交通大学 | Man-machine dialogue anomaly detection system and method |
| CN107278302A (en) * | 2017-03-02 | 2017-10-20 | 深圳前海达闼云端智能科技有限公司 | A kind of robot interactive method and interaction robot |
| CN109119084A (en) * | 2018-07-13 | 2019-01-01 | 广东电网有限责任公司 | A kind of dispatch call method and system based on speech recognition |
| CN109413286A (en) * | 2018-10-22 | 2019-03-01 | 北京移数通电讯有限公司 | A kind of intelligent customer service voice response system and method |
Non-Patent Citations (1)
| Title |
|---|
| 朱频频: "《智能客户服务技术与应用》", 30 January 2019 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992941A (en) * | 2019-10-22 | 2020-04-10 | 国网天津静海供电有限公司 | Power grid dispatching voice recognition method and device based on spectrogram |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107222865B (en) | Communication swindle real-time detection method and system based on suspicious actions identification | |
| US10847146B2 (en) | Multiple voice recognition model switching method and apparatus, and storage medium | |
| CN110334241B (en) | Quality inspection method, device and equipment for customer service record and computer readable storage medium | |
| CN101751922B (en) | Text-independent speech conversion system based on HMM model state mapping | |
| CN103458056B (en) | Speech intention judging system based on automatic classification technology for automatic outbound system | |
| CN105374356B (en) | Audio recognition method, speech assessment method, speech recognition system and speech assessment system | |
| CN101447185B (en) | Audio frequency rapid classification method based on content | |
| CA3176868A1 (en) | Intent identifying method and device for application to intelligent customer service robot | |
| CN109165284A (en) | A kind of financial field human-computer dialogue intension recognizing method based on big data | |
| CN110209791A (en) | It is a kind of to take turns dialogue intelligent speech interactive system and device more | |
| CN107680602A (en) | Voice fraud recognition methods, device, terminal device and storage medium | |
| CN110335609A (en) | A kind of air-ground communicating data analysis method and system based on speech recognition | |
| CN106935239A (en) | The construction method and device of a kind of pronunciation dictionary | |
| CN105261367B (en) | A kind of method for distinguishing speek person | |
| CN109308466A (en) | The method that a kind of pair of interactive language carries out Emotion identification | |
| CN106297826A (en) | Speech emotional identification system and method | |
| CN107305773B (en) | Voice emotion recognition method | |
| CN110517664A (en) | Multi-party speech recognition methods, device, equipment and readable storage medium storing program for executing | |
| CN105654940B (en) | Speech synthesis method and device | |
| CN103065620A (en) | Method with which text input by user is received on mobile phone or webpage and synthetized to personalized voice in real time | |
| US20210049996A1 (en) | Voice recognition method using artificial intelligence and apparatus thereof | |
| CN111324708A (en) | Natural language processing system based on human-computer interaction | |
| CN106356054A (en) | Method and system for collecting information of agricultural products based on voice recognition | |
| CN114627863A (en) | Speech recognition method and device based on artificial intelligence | |
| CN113609264B (en) | Data query method and device for power system nodes |
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: 20190920 |
|
| RJ01 | Rejection of invention patent application after publication |