CN106971710A - Electricity business hall personnel voice anomalous event recognition methods and device - Google Patents
Electricity business hall personnel voice anomalous event recognition methods and device Download PDFInfo
- Publication number
- CN106971710A CN106971710A CN201710154029.9A CN201710154029A CN106971710A CN 106971710 A CN106971710 A CN 106971710A CN 201710154029 A CN201710154029 A CN 201710154029A CN 106971710 A CN106971710 A CN 106971710A
- Authority
- CN
- China
- Prior art keywords
- voice
- short
- time energy
- module
- speech
- 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
- 230000002547 anomalous effect Effects 0.000 title claims abstract description 41
- 230000005611 electricity Effects 0.000 title claims abstract description 18
- 238000000034 method Methods 0.000 title claims abstract description 15
- 230000002159 abnormal effect Effects 0.000 claims abstract description 48
- 238000001514 detection method Methods 0.000 claims abstract description 14
- 238000012545 processing Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 5
- 230000005856 abnormality Effects 0.000 claims description 4
- 238000013500 data storage Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000000605 extraction Methods 0.000 claims 1
- 238000007726 management method Methods 0.000 abstract description 7
- 230000008451 emotion Effects 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 abstract description 4
- 206010027940 Mood altered Diseases 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 230000002996 emotional effect Effects 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000007510 mood change Effects 0.000 abstract description 3
- 230000036651 mood Effects 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000012544 monitoring process Methods 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
- G10L15/04—Segmentation; Word boundary detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06398—Performance of employee with respect to a job function
-
- 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
- G10L15/04—Segmentation; Word boundary detection
- G10L15/05—Word boundary detection
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
- G10L25/63—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for estimating an emotional state
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Strategic Management (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Hospice & Palliative Care (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Child & Adolescent Psychology (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Telephonic Communication Services (AREA)
Abstract
The present invention relates to service level assessment technology field, specifically a kind of electricity business hall personnel voice anomalous event recognition methods and device based on short-time energy phonetic feature, including：Voice pretreatment module, the main voice to input carries out the pretreatment before use, voice pretreatment module includes endpoint detection module and denoising module, wherein endpoint detection module is to be detected using voice short-time average energy at the end points of voice, and denoising module carries out denoising to voice；The present invention being capable of intellectual analysis shop assistant's voice, it can note controlling current emotional with intelligent reminding shop assistant when shop assistant's mood changes, using preferable attitude as customer service, increase customer satisfaction degree, to shop assistant's abnormal emotion archive management, reference can be improved for shop assistant's assessment of performance, play good supervision and management effect.
Description
Technical field
The present invention relates to service level assessment technology field, specifically a kind of electricity based on short-time energy phonetic feature
The voice anomalous event recognition methods of power business hall person and device.
Background technology
The service quality of enterprise is related to the development of enterprise, if service is made not in place, cannot Win Clients and whole
The public praise of individual industry, not good public praise enterprise will be unable to tremendous development.So, the service quality of enterprise just seems heavy to closing
Will.The service window of electricity consumption business is handled in the Power supply business Room as power customer, is directly come into contacts with face-to-face with client, and work people
Every word and action of member shows Enterprise Quality of Service and state's net brand image invariably.The quality of business personnel, which is directly affected, powers
Service quality, as the business personnel that region be directly facing customer service, the quality of its attitude is related to customer experience and service
Quality.Therefore, the supervision to business personnel's attitude is realized, one of effective means for lifting its service quality can be used as.
Attitude supervision is carried out to business personnel, the means that can be used mainly there are two kinds：The first is collection video
Information, according to video monitoring, shop assistant is judged by information such as the facial expression of shop assistant, actions by way of manually supervising
Attitude, this mode needs manually to stand fast at, and adds cost of labor, and due to the weakness on people's own physiological, nothing
Method discovers tiny plot, so as to cause managerial careless omission so that the efficiency of management is than relatively low.The second way is collection business
The voice dialogue of member, the current emotional state of shop assistant is judged by analyzing voice, this mode is realized simply, without artificial
Stand fast at, training in rotation, system automatically analyzes push abnormal information.
When being supervised to shop assistant's mood, analyzed using intelligent audio, it is rich with directly perceived, accurate, timely and content
The superiority such as richness.Short-time energy in voice shows as the loudness of sound, the sound of sound when people is in the excitement mood such as indignation
Degree can increase a lot, therefore utilize the short-time energy characteristic value of voice to judge the emotional change of shop assistant, with good differentiation
Property.
The content of the invention
To solve the above problems, proposing a kind of electricity business hall personnel's voice anomalous event identification side based on short-time energy
Method and device, by the speech analysis to electric power shop assistant, monitor shop assistant's attitude, when abnormal emotion occurs in shop assistant
Abnormal emotion event is achieved, the reference of examination shop assistant is used as.
To achieve the above object, the present invention uses following scheme：
A kind of electricity business hall personnel voice anomalous event recognition methods, is comprised the steps of：
Step A, is pre-processed to the voice of input, mainly carries out end-point detection and denoising to voice, then
Voice after output processing,
Step B, pretreated voice is split, and intercepts into size identical voice segments, calculates each voice segments
Short-time energy characteristic value, export each voice segments and its short-time energy characteristic value,
The data of step C, receiving step B output, using the short-time energy characteristic value of first paragraph voice segments as reference point, its
He is compared the short-time energy characteristic value of voice segments with reference point respectively, judges the abnormal conditions of each voice segments, exports different
Normal voice segments,
Step D, handles anomalous event, carries out data storage to the abnormal speech that step C is exported, abnormal events information is pushed away
Keeper is given to be examined.
In the step A, the end-point detection to voice is mainly included the following steps that：
Step A1, the excessive threshold value of short-time average energy of unvoiced speech section and speech sound section is trained using iterative algorithm,
Step A2, sub-frame processing voice to be detected extracts the first frame of voice,
Step A3, calculates the short-time average energy value for having extracted frame, judges whether to exceed excessive threshold value, is not above, carries
Next frame is taken, step A3 is continued executing with, more than the end points of voice segments when then illustrating the frame, step A4 is performed,
Step A4, exports the voice after the end points detected.
Iterative algorithm in the step A1, implements step as follows：
A1.1 collecting quantity identicals unvoiced speech section and speech sound section training sample,
A1.2 calculates the short-time energy characteristic value of each voice segments, foundation
Calculate, wherein w (n) is window function, N is that window is long,
A1.3 asks the average short-time energy value M1 of unvoiced speech section and the average short-time energy value M2 of speech sound section, setting
M=(M1+M2)/2 is initial threshold,
All test samples of A1.4, if its short-time energy value be determined as more than threshold value M it is sound, be otherwise determined as it is noiseless,
A1.5 calculates the accuracy rate of the speech sound judged and unvoiced speech according to judged result and sample data,
If the accuracy rate of A1.6 unvoiced speech is more than the accuracy rate of speech sound, downwards adjustment threshold value M, perform
A1.4, if the accuracy rate that the accuracy rate of unvoiced speech is less than speech sound adjusts upward threshold value M, performs A1.4.When noiseless language
The accuracy rate of sound then returns to threshold value M for speech sound and unvoiced speech excessive threshold value when identical more than the accuracy rate of speech sound.
In the step B, pretreated voice is split, size identical voice segments are intercepted into, calculates each
The short-time energy characteristic value of voice segments, exports each voice segments and its short-time energy characteristic value, mainly realizes that step is as follows：
B1, period identical voice segments are divided into by pretreated voice, should by the initial time name of voice segments
Voice segments,
B2, calculates the short-time energy characteristic value of every section of voice, passes throughMeter
Calculate, wherein w (n) is window function, N is that window is long,
B3, each voice segments and its short-time energy characteristic value are exported,
In the step C, the data of receiving step B outputs regard the short-time energy characteristic value of first paragraph voice segments as ginseng
According to value, the short-time energy characteristic value of other voice segments is compared with reference point respectively, judges the abnormal conditions of each voice segments,
Output abnormality voice segments.Key step is as follows：
C1, in the voice segments that B2 is inputted, chooses the short-time energy characteristic value of wherein first voice segments of time earliest
The reference point whether abnormal as voice segments are judged,
C2, the short-time energy characteristic values of other each voice segments and the reference point chosen do division to a ratio.
C3, sets a threshold value, if the ratio obtained in C2 is less than the threshold value, voice segments are normal, otherwise the voice
Duan Yichang, abnormal speech is exported,
The setting of threshold value in wherein C3, is that the average value of the short-time energy value of a large amount of angry voice segments of basis is put down with a large amount of
The average value of the short-time energy value of quiet voice segments is done than being worth to.
In the step D, anomalous event is handled, data storage is carried out to the abnormal speech that step C is exported, by anomalous event
Information is pushed to keeper and examined.Comprise the following steps that：
D1, by the abnormal speech exported in C section deposit distributed file system, returns to the address of abnormal speech section,
D2, address of the abnormal speech section in distributed file system is stored in local data base,
D3, is pushed to keeper by the anomalous event and is examined, keeper gets abnormal language according to database address
Whether segment is examined abnormal.
Business hall person voice anomalous event identifying device based on short-time energy, including：
Voice pretreatment module, mainly carries out the pretreatment before use, voice pretreatment module is included to the voice of input
Endpoint detection module and denoising module, wherein endpoint detection module are to detect voice using voice short-time average energy
At end points, denoising module carries out denoising to voice；
Characteristic extracting module, mainly realizes the calculating of voice short-time energy characteristic value, at the sound end continuous cut
The voice segments of equal length are taken, the short-time energy value of each voice segments is calculated；
Anomalous event judge module, the short-time energy value of each voice segments drawn according to characteristic extracting module, judging should
The two states of voice segments, set a threshold value, if the short-time energy value of voice is more than the threshold value, are determined as abnormal shape
State, otherwise, it is determined that being normal condition；
Anomalous event processing module, the anomalous event that the module is inputted according to anomalous event judge module is responded, will
The abnormal speech is uploaded to document storage system module, and voice address is stored in local data base；
Document storage system module, the module uses distributed file storage system, for storing abnormal speech section；
Database module, local data base is used to store address of the abnormal speech section in document storage system.
Beneficial effects of the present invention：
Intellectual analysis shop assistant's voice, can notice that control is worked as when shop assistant's mood changes with intelligent reminding shop assistant
Preceding mood, using preferable attitude as customer service, increases customer satisfaction degree.
To shop assistant's abnormal emotion archive management, reference can be improved for shop assistant's assessment of performance, play good supervision
Management effect.
Brief description of the drawings：
Electricity business hall personnel voice anomalous event identifying devices of the Fig. 1 based on short-time energy
Electricity business hall personnel voice anomalous event recognition methods flow charts of the Fig. 2 based on short-time energy
Embodiment：
The invention will be further described with embodiment below in conjunction with the accompanying drawings.
As shown in figure 1, be electricity business hall personnel's voice anomalous event identifying device based on short-time energy, main bag
Include：
Voice pretreatment module, the voice of input is pre-processed, and detects the end points of voice, and to the voice of input
Carry out denoising.Voice pretreatment module includes endpoint detection module and denoising module, and endpoint detection module is to utilize voice
End points of the short-time average energy value to determine voice at；Denoising module is to remove the noise in voice；
Characteristic extracting module, the module realizes the calculating to the short-time energy characteristic value of voice segments, and voice is carried out first
Isometric dividing processing, intercepts the voice segments of certain length, then calculates the short-time energy value of the voice segments, and give anomalous event
Judge module judges the state of the voice；
Anomalous event judge module, the short-time energy value of the voice segments drawn according to characteristic extracting module, judges the voice
The state of section, point normally with abnormal two states.The first paragraph most started using voice is as reference, because being sought during beginning of conversation
Industry person's mood is all that comparison is tranquil, and then each section of voice segments are contrasted with first paragraph voice segments, if ratio exceedes setting threshold
Value, then be judged as exception.The threshold value is the tranquil language of the average short-time energy value using a large amount of angry voice segments and identical quantity
The average short-time energy value of segment does what ratio was drawn；
Anomalous event processing module, the anomalous event that the module is inputted according to anomalous event judge module is responded, will
The abnormal speech is uploaded to document storage system module, and voice address is stored in local data base；
Document storage system module, the module uses distributed file storage system, for storing abnormal speech section, is needing
It can be recalled at any time when checking.With the growth of data volume, distributed memory system, which can be very good reply storage, to be held
The problem of small, data of amount increase in terms of fast, data backup, data safety；
Database module, local data base is used to store address of the abnormal speech section in document storage system, is carrying out
During abnormal speech section inquiry, voice segments are found in the address of document storage system in local data base first, then base area
Voice segments content is checked in location.
As shown in Fig. 2 be electricity business hall personnel's voice anomalous event recognition methods flow chart based on short-time energy, stream
Journey is as follows：
Step 101, the voice of input is pre-processed, end-point detection and denoising mainly is carried out to voice, so
Voice after output is handled afterwards.End-point detection：First with iterative algorithm training unvoiced speech section and speech sound section in short-term
The excessive threshold value of average energy, 50 unvoiced speech sections of collection and 50 speech sound section training samples, calculate each voice segments
Short-time energy characteristic value, asks the average short-time energy value M1 of unvoiced speech section and the average short-time energy value M2 of speech sound section,
It is initial threshold to set M=(M1+M2)/2.All test samples are judged, if its short-time energy value is determined as having more than threshold value M
Sound, is otherwise determined as noiseless.The accuracy rate of the speech sound judged and unvoiced speech is calculated, if the accuracy rate of unvoiced speech is big
In the accuracy rate of speech sound, then threshold value M is adjusted downwards, continue judgement sample, if the accuracy rate of unvoiced speech is less than sound
The accuracy rate of voice adjusts upward threshold value M, continues judgement sample.When the accuracy rate of unvoiced speech is more than the accuracy rate of speech sound
Threshold value M is then returned when identical for speech sound and the excessive threshold value of unvoiced speech.Then sub-frame processing voice to be detected, extracts language
All frames of sound and the short-time average energy value for calculating all frames, judge whether to exceed excessive threshold value frame by frame since the first frame,
More than then illustrating that the frame is the end points of voice segments, the voice after the end points detected is exported.
Step 102, pretreated voice is split, intercepts the voice segments of 5s length successively by phoneme sequence, and with
The voice segments time started is named, and calculates the short-time energy characteristic value of each voice segments, exports each voice segments and its short-time energy
Characteristic value.
Step 103, the data that receiving step 102 is exported, regard the short-time energy characteristic value of first paragraph voice segments as reference
Value, the short-time energy characteristic value of other voice segments does division with reference point progress respectively and obtains a ratio, sets a threshold value,
The threshold value is according to the average value of the short-time energy value of 50 angry voice segments and the short-time energy value of 50 tranquil voice segments
Average value do than being worth to.If obtained ratio is less than the threshold value, voice segments are normal, and otherwise the voice segments are abnormal,
Abnormal speech is exported.
Step 104, anomalous event is handled, distributed file system is stored in the abnormal speech that step 103 is exported, returning should
The address of abnormal speech section.The address is stored in local data base.The anomalous event is pushed into keeper to be examined, managed
Reason person gets abnormal speech section according to database address and whether abnormal examines.
Beneficial effects of the present invention：
Intellectual analysis shop assistant's voice, can notice that control is worked as when shop assistant's mood changes with intelligent reminding shop assistant
Preceding mood, using preferable attitude as customer service, increases customer satisfaction degree.
To shop assistant's abnormal emotion archive management, reference can be improved for shop assistant's assessment of performance, play good supervision
Management effect.
Claims (7)
1. a kind of electricity business hall personnel voice anomalous event recognition methods, it is characterised in that comprise the steps of：
Step A, is pre-processed to the voice of input, is mainly carried out end-point detection and denoising to voice, is then exported
Voice after processing,
Step B, pretreated voice is split, and intercepts into size identical voice segments, calculates the short of each voice segments
When energy eigenvalue, export each voice segments and its short-time energy characteristic value,
The data of step C, receiving step B output, regard the short-time energy characteristic value of first paragraph voice segments as reference point, other languages
The short-time energy characteristic value of segment is compared with reference point respectively, judges the abnormal conditions of each voice segments, output abnormality language
Segment,
Step D, handles anomalous event, carries out data storage to the abnormal speech that step C is exported, abnormal events information is pushed to
Keeper is examined.
2. a kind of electricity business hall personnel voice anomalous event recognition methods according to claim 1, it is characterised in that institute
State in step A, the end-point detection to voice is mainly included the following steps that：
Step A1, the excessive threshold value of short-time average energy of unvoiced speech section and speech sound section is trained using iterative algorithm,
Step A2, sub-frame processing voice to be detected extracts the first frame of voice,
Step A3, calculates the short-time average energy value for having extracted frame, judges whether to exceed excessive threshold value, is not above, under extraction
One frame, continues executing with step A3, more than the end points of voice segments when then illustrating the frame, performs step A4,
Step A4, exports the voice after the end points detected.
3. a kind of electricity business hall personnel voice anomalous event recognition methods according to claim 2, it is characterised in that step
Iterative algorithm in rapid A1, implements step as follows：
A1.1 collecting quantity identicals unvoiced speech section and speech sound section training sample,
A1.2 calculates the short-time energy characteristic value of each voice segments, foundationMeter
Calculate, wherein w (n) is window function, N is that window is long,
A1.3 asks the average short-time energy value M1 of unvoiced speech section and the average short-time energy value M2 of speech sound section, sets M=
(M1+M2)/2 it is initial threshold,
All test samples of A1.4, if its short-time energy value be determined as more than threshold value M it is sound, be otherwise determined as it is noiseless,
A1.5 calculates the accuracy rate of the speech sound judged and unvoiced speech according to judged result and sample data,
If the accuracy rate of A1.6 unvoiced speech is more than the accuracy rate of speech sound, downwards adjustment threshold value M, A1.4 is performed, such as
The accuracy rate that the accuracy rate of fruit unvoiced speech is less than speech sound adjusts upward threshold value M, performs A1.4.It is accurate when unvoiced speech
Rate then returns to threshold value M for speech sound and unvoiced speech excessive threshold value when identical more than the accuracy rate of speech sound.
4. a kind of electricity business hall personnel voice anomalous event recognition methods according to claim 1, it is characterised in that institute
State in step B, pretreated voice is split, intercept into size identical voice segments, calculate the short of each voice segments
When energy eigenvalue, export each voice segments and its short-time energy characteristic value, mainly realize that step is as follows：
B1, is divided into period identical voice segments by pretreated voice, the voice is named by the initial time of voice segments
Section,
B2, calculates the short-time energy characteristic value of every section of voice, passes throughCalculate, its
Middle w (n) is window function, and N is that window is long,
B3, each voice segments and its short-time energy characteristic value are exported.
5. a kind of electricity business hall personnel voice anomalous event recognition methods according to claim 1, it is characterised in that institute
State in step C, the data of receiving step B outputs regard the short-time energy characteristic value of first paragraph voice segments as reference point, other languages
The short-time energy characteristic value of segment is compared with reference point respectively, judges the abnormal conditions of each voice segments, output abnormality language
Segment.Key step is as follows：
C1, in the voice segments that B2 is inputted, chooses the short-time energy characteristic value conduct of wherein first voice segments of time earliest
Judge the whether abnormal reference point of voice segments,
C2, the short-time energy characteristic values of other each voice segments and the reference point chosen do division to a ratio.
C3, sets a threshold value, if the ratio obtained in C2 is less than the threshold value, voice segments are normal, and otherwise the voice segments are different
Often, abnormal speech is exported,
The setting of threshold value in wherein C3, is the average value and a large amount of calmness of the short-time energy value of a large amount of angry voice segments of basis
The average value of the short-time energy value of voice segments is done than being worth to.
6. a kind of electricity business hall personnel voice anomalous event recognition methods according to claim 1, it is characterised in that institute
State in step D, handle anomalous event, data storage is carried out to the abnormal speech that step C is exported, abnormal events information is pushed to
Keeper is examined.Comprise the following steps that：
D1, by the abnormal speech exported in C section deposit distributed file system, returns to the address of abnormal speech section,
D2, address of the abnormal speech section in distributed file system is stored in local data base,
D3, is pushed to keeper by the anomalous event and is examined, and keeper gets abnormal speech section according to database address
Whether abnormal examine.
7. a kind of electricity business hall personnel voice anomalous event identifying device, it is characterised in that people from business hall based on short-time energy
Member's voice anomalous event identifying device, including：
Voice pretreatment module, mainly carries out the pretreatment before use, voice pretreatment module includes end points to the voice of input
Detection module and denoising module, wherein endpoint detection module are the end points that voice is detected using voice short-time average energy
Place, denoising module carries out denoising to voice；
Characteristic extracting module, mainly realizes the calculating of voice short-time energy characteristic value, at the sound end continuous interception phase
With the voice segments of length, the short-time energy value of each voice segments is calculated；
Anomalous event judge module, the short-time energy value of each voice segments drawn according to characteristic extracting module, judges the voice
The two states of section, set a threshold value, if the short-time energy value of voice is more than the threshold value, are determined as abnormality, no
Then, it is determined as normal condition；
Anomalous event processing module, the anomalous event that the module is inputted according to anomalous event judge module is responded, and this is different
Chang Yuyin is uploaded to document storage system module, and voice address is stored in local data base；
Document storage system module, the module uses distributed file storage system, for storing abnormal speech section；
Database module, local data base is used to store address of the abnormal speech section in document storage system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710154029.9A CN106971710A (en) | 2017-03-15 | 2017-03-15 | Electricity business hall personnel voice anomalous event recognition methods and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710154029.9A CN106971710A (en) | 2017-03-15 | 2017-03-15 | Electricity business hall personnel voice anomalous event recognition methods and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106971710A true CN106971710A (en) | 2017-07-21 |
Family
ID=59329609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710154029.9A Pending CN106971710A (en) | 2017-03-15 | 2017-03-15 | Electricity business hall personnel voice anomalous event recognition methods and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106971710A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110705505A (en) * | 2019-10-14 | 2020-01-17 | 国网山东省电力公司泰安供电公司 | Intelligent identification device, method and system for service specification of electric power business hall |
| CN113838478A (en) * | 2020-06-08 | 2021-12-24 | 华为技术有限公司 | Abnormal event detection method and device and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080195385A1 (en) * | 2007-02-11 | 2008-08-14 | Nice Systems Ltd. | Method and system for laughter detection |
| CN102509546A (en) * | 2011-11-11 | 2012-06-20 | 北京声迅电子股份有限公司 | Noise reduction and abnormal sound detection method applied to rail transit |
| CN102610228A (en) * | 2011-01-19 | 2012-07-25 | 上海弘视通信技术有限公司 | Audio exception event detection system and calibration method for the same |
| CN103198838A (en) * | 2013-03-29 | 2013-07-10 | 苏州皓泰视频技术有限公司 | Abnormal sound monitoring method and abnormal sound monitoring device used for embedded system |
| CN103267568A (en) * | 2013-05-29 | 2013-08-28 | 哈尔滨工业大学 | Voice online detection method for automobile electronic control unit |
| CN106128475A (en) * | 2016-07-12 | 2016-11-16 | 华南理工大学 | Wearable intelligent safety equipment based on abnormal emotion speech recognition and control method |
-
2017
- 2017-03-15 CN CN201710154029.9A patent/CN106971710A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080195385A1 (en) * | 2007-02-11 | 2008-08-14 | Nice Systems Ltd. | Method and system for laughter detection |
| CN102610228A (en) * | 2011-01-19 | 2012-07-25 | 上海弘视通信技术有限公司 | Audio exception event detection system and calibration method for the same |
| CN102509546A (en) * | 2011-11-11 | 2012-06-20 | 北京声迅电子股份有限公司 | Noise reduction and abnormal sound detection method applied to rail transit |
| CN103198838A (en) * | 2013-03-29 | 2013-07-10 | 苏州皓泰视频技术有限公司 | Abnormal sound monitoring method and abnormal sound monitoring device used for embedded system |
| CN103267568A (en) * | 2013-05-29 | 2013-08-28 | 哈尔滨工业大学 | Voice online detection method for automobile electronic control unit |
| CN106128475A (en) * | 2016-07-12 | 2016-11-16 | 华南理工大学 | Wearable intelligent safety equipment based on abnormal emotion speech recognition and control method |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110705505A (en) * | 2019-10-14 | 2020-01-17 | 国网山东省电力公司泰安供电公司 | Intelligent identification device, method and system for service specification of electric power business hall |
| CN113838478A (en) * | 2020-06-08 | 2021-12-24 | 华为技术有限公司 | Abnormal event detection method and device and electronic equipment |
| CN113838478B (en) * | 2020-06-08 | 2024-04-09 | 华为技术有限公司 | Abnormal event detection method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107818798B (en) | Customer service quality evaluation method, device, equipment and storage medium | |
| CN108122552A (en) | Voice mood recognition methods and device | |
| Mohan | Speech recognition using MFCC and DTW | |
| CN102623009B (en) | Abnormal emotion automatic detection and extraction method and system on basis of short-time analysis | |
| CN110570873B (en) | Voiceprint wake-up method and device, computer equipment and storage medium | |
| CN111080109B (en) | Customer service quality evaluation method and device and electronic equipment | |
| CN103258535A (en) | Identity recognition method and system based on voiceprint recognition | |
| CN110930989B (en) | Speech intention recognition method and device, computer equipment and storage medium | |
| Ghai et al. | Emotion recognition on speech signals using machine learning | |
| Mporas et al. | Estimation of unknown speaker’s height from speech | |
| WO2017045429A1 (en) | Audio data detection method and system and storage medium | |
| CN106971710A (en) | Electricity business hall personnel voice anomalous event recognition methods and device | |
| CN112185385A (en) | Intelligent client processing method and system for power supply field | |
| CN113436634A (en) | Voice classification method and device based on voiceprint recognition and related equipment | |
| Yan et al. | Speech interactive emotion recognition system based on random forest | |
| Pham et al. | Toward better speaker embeddings: Automated collection of speech samples from unknown distinct speakers | |
| CN111489736B (en) | Automatic scoring device and method for seat speaking operation | |
| Verkholyak et al. | Hierarchical two-level modelling of emotional states in spoken dialog systems | |
| Gadhe et al. | Emotion recognition from speech: a survey | |
| CN115063155B (en) | Data labeling method, device, computer equipment and storage medium | |
| CN110852178A (en) | Piano music score difficulty identification method based on decision tree lifting | |
| CN103531208B (en) | A kind of space flight stress emotion identification method based on short term memory weight fusion | |
| JP7096296B2 (en) | Information processing equipment, information processing methods and programs | |
| Warule et al. | Hilbert-Huang Transform-Based Time-Frequency Analysis of Speech Signals for the Identification of Common Cold | |
| Sharma et al. | Speech Emotion Recognition System using SVD algorithm with HMM Model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170721 |