CN116665708A - Illegal service operation detection system and method thereof - Google Patents
Illegal service operation detection system and method thereof Download PDFInfo
- Publication number
- CN116665708A CN116665708A CN202211099629.7A CN202211099629A CN116665708A CN 116665708 A CN116665708 A CN 116665708A CN 202211099629 A CN202211099629 A CN 202211099629A CN 116665708 A CN116665708 A CN 116665708A
- Authority
- CN
- China
- Prior art keywords
- voiceprint
- voice
- user
- offline
- incoming call
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000008569 process Effects 0.000 claims abstract description 24
- 238000007781 pre-processing Methods 0.000 claims description 31
- 238000005065 mining Methods 0.000 claims description 27
- 238000000605 extraction Methods 0.000 claims description 18
- 238000012795 verification Methods 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 9
- 238000000926 separation method Methods 0.000 claims description 6
- 238000007619 statistical method Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 10
- 230000001960 triggered effect Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011897 real-time detection Methods 0.000 description 2
- 238000012550 audit Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 230000008520 organization 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
- 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
-
- 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
- G10L17/02—Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/08—Network architectures or network communication protocols for network security for authentication of entities
- H04L63/0861—Network architectures or network communication protocols for network security for authentication of entities using biometrical features, e.g. fingerprint, retina-scan
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/32—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols including means for verifying the identity or authority of a user of the system or for message authentication, e.g. authorization, entity authentication, data integrity or data verification, non-repudiation, key authentication or verification of credentials
- H04L9/3226—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols including means for verifying the identity or authority of a user of the system or for message authentication, e.g. authorization, entity authentication, data integrity or data verification, non-repudiation, key authentication or verification of credentials using a predetermined code, e.g. password, passphrase or PIN
- H04L9/3228—One-time or temporary data, i.e. information which is sent for every authentication or authorization, e.g. one-time-password, one-time-token or one-time-key
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a illegal service operation detection system and a method thereof, wherein the system comprises an application end and a voiceprint device which are connected with each other, the application end transmits voice data or offline voice files of a future user to the voiceprint device, the voiceprint device analyzes and processes the voice data or offline voice files of the future user so as to output and obtain corresponding detection results, and the detection results and corresponding instructions are transmitted to the application end; and the other party carries out grouping classification comparison on mass voice files in a certain period according to the talkers, and then further digs potential illegal users by combining with the business rules. Therefore, the detection of illegal business operation can be accurately and reliably realized, potential illegal personnel can be excavated in advance, and corresponding early warning can be carried out.
Description
Technical Field
The invention relates to the technical field of illegal service detection, in particular to an operation detection system and method for illegal service.
Background
In the existing business scene, due to the fact that identity card information is revealed, illegal personnel steal the revealed identity card information and carry out false beautifying processing on histories and backgrounds of the identity cards so as to cheat business audit, and the illegal personnel can obtain improper benefits from the records, thereby bringing bad accounts and great losses to financial institutions.
In general, illegal personnel are very familiar with the flow of related business auditing, and thus, no valuable clues are left in the whole business operation process except for the call record with the business staff. This results in the fact that after an illegal event, the institutional party has difficulty finding and giving interference deterrence at the first time. Often, institutions are aware of the occurrence of financial fraud after the occurrence of an incumbent malicious fact. And even after a certain fraudulent individual case is exposed, the organization side can only trace and manage the individual case at present, can not carry out all tracing and management and control on all fraudulent cases related to the illegal personnel, can not thoroughly eliminate all risks related to the illegal personnel, and can not early warn risks brought by other similar illegal personnel.
That is, in the face of fraudulent activity of an illegal person, the prior art can only take measures for risk management and control for individual cases after a fact, and cannot mine potential illegal activity of an illegal person in advance.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a system and a method for detecting illegal business operation, which can pre-mine potential illegal personnel and perform corresponding early warning.
The aim of the invention can be achieved by the following technical scheme: the illegal service operation detection system comprises an application end and a voiceprint device which are connected with each other, wherein the application end transmits voice data or offline voice files of a calling user to the voiceprint device, and the voiceprint device analyzes and processes the voice data or offline voice files of the calling user so as to output corresponding detection results and transmits the detection results and corresponding instructions to the application end.
Further, the application end is provided with a recording acquisition device for collecting voice data of the incoming call user.
Further, the voiceprint device comprises an online voiceprint unit and an offline voiceprint unit, wherein the online voiceprint unit is used for carrying out online real-time illegal operation detection, and the offline voiceprint unit is used for carrying out illegal operation detection on offline massive voices.
Further, the online voiceprint unit comprises a preprocessing module, a voiceprint feature extraction module, a voiceprint registration module, a verification comparison module and a database, wherein the preprocessing module is used for removing non-user sounds in incoming call user voice data; the voiceprint feature extraction module is used for extracting voiceprint features of the caller from the preprocessed caller voice data; the voiceprint registration module is used for storing voiceprint features of the incoming call user in a database and establishing a unique mapping relation between the voiceprint features of the incoming call user and a user Identity (ID); the verification comparison module is used for verifying and comparing the voice print characteristics of the incoming call user with voice prints corresponding to the specific list library and the user identity ID so as to output corresponding detection results; and a specific list library and a user identity ID-voiceprint library are arranged in the database.
Further, the offline voiceprint unit comprises a preprocessing module, a batch voiceprint registration module, a batch voiceprint comparison module, a clustering module, an excavating module and a database, wherein the preprocessing module is used for removing non-user sounds in an offline voice file; the batch voiceprint registration module is used for carrying out offline batch scanning, batch extraction and batch registration on the preprocessed offline voice files, and storing voiceprint characteristics corresponding to file names in a database; the batch voiceprint comparison module is used for carrying out pairwise combination comparison on voiceprints which are registered in batch, and outputting to obtain voiceprint comparison scores corresponding to a plurality of combinations; the clustering module outputs a caller grouping result by adopting a merging and classifying mode according to the voiceprint comparison scores; and the mining module outputs a specific list according to the grouping result of the talkers and the set mining strategy, and stores the specific list in the database.
An on-line detection method for illegal service operation comprises the following steps:
a1, transmitting incoming user voice data to an online voiceprint unit in real time by an application end;
a2, after the online voiceprint unit receives the voice data of the incoming call user, voice preprocessing, voiceprint feature extraction, voiceprint registration and voiceprint verification comparison operation are sequentially carried out, and an online detection result of the illegal operation is obtained and transmitted to the application end.
Further, the step A2 specifically includes the following steps:
a21, preprocessing the voice data of the incoming call user to remove non-user voice, wherein the preprocessing process comprises silence detection, voice quality detection, noise removal and speaker separation;
a22, extracting voice print characteristics of the incoming call user from the preprocessed voice data of the incoming call user based on the deep neural network;
a23, storing the extracted voiceprint features of the incoming call user into a database in a binary unreadable form, and establishing a unique mapping relation between the voiceprint features of the incoming call user and the user identity ID so as to finish online voiceprint registration;
a24, comparing the voiceprint features of the incoming call user with the voiceprint features in the specific list library, and outputting alarm information to the application end if the voiceprint features of the incoming call user are judged to belong to the specific list;
otherwise, comparing the voiceprint features of the incoming call user with the voiceprint features of the corresponding user Identity (ID) in the database, and outputting a corresponding comparison result to the application end.
An off-line detection method for illegal service operation comprises the following steps:
b1, the application end transmits the collected offline voice file in the set time period to an offline voiceprint unit;
and B2, after the offline voiceprint unit receives the offline voice file, carrying out voice preprocessing, batch voiceprint registration, batch voiceprint comparison, speaker clustering and mining operation in sequence, and outputting to obtain specific list data.
Further, the step B2 specifically includes the following steps:
b21, preprocessing the offline voice file to remove non-user sounds, wherein the preprocessing process comprises silence detection, voice quality detection, noise removal and speaker separation;
b22, carrying out batch scanning and batch extraction on the preprocessed offline voice file, then storing the extracted voice characteristic into a database in a binary unreadable form, and establishing a unique mapping relation between the voice characteristic and a corresponding offline voice file name to sequentially finish offline batch voice registration;
b23, carrying out pairwise combination comparison on offline voiceprints which are registered in batches to obtain a plurality of combination voiceprint comparison scores;
b24, according to a plurality of combined voiceprint comparison score results, combining and classifying voices belonging to different speakers, and outputting to obtain speaker grouping results;
and B25, outputting and obtaining specific list data according to a caller grouping result and a set mining strategy, wherein the mining strategy comprises a number strategy, an ID strategy, a calling frequency strategy and an area strategy.
Further, the number policy specifically includes: for each caller group file, scanning caller number information of all voice files in the group, if the voice files in the group come from more than 2 different numbers, putting the caller into a suspected specific list;
the ID strategy specifically comprises the following steps: for each speaker group file, scanning user ID information of all voice files in the group, and if the voice files in the group come from more than 2 different user IDs, putting the speaker into a suspected specific list;
the calling frequency strategy specifically comprises the following steps: aiming at the files grouped by each caller, scanning the incoming call time of all voice files in the group, carrying out statistical analysis on the incoming call time of each voice file, converting the incoming call time into calling frequency, and listing the callers corresponding to the calling frequency higher than the set calling frequency threshold value in a suspected specific list;
the regional policy specifically includes: and acquiring incoming call region information of all the voice files from the suspected specific list result, carrying out statistical analysis on the region information, calculating the duty ratio, and if the duty ratio of a certain region exceeds a set region duty ratio threshold value, listing the region as a risk region and outputting the result.
Compared with the prior art, the voice-print device comprises the online voice-print unit and the offline voice-print unit, the online voice-print unit is used for detecting on-line real-time illegal operation, and the offline voice-print unit is used for detecting off-line massive voice. On one hand, the user voice and the specific list voiceprint library are compared in real time on line, and a result is returned in time; and the offline voice files of all the callers in a certain period are grouped, classified and compared according to the callers under the line of the other party, and then the potential specific list data is further mined by combining the business rules.
The invention provides an online detection method aiming at real-time incoming call user voice data, which can reliably compare and detect the user incoming call voice data with voiceprints in a specific list library by preprocessing the user incoming call voice data, extracting voiceprint features, registering voiceprints and verifying and comparing the voiceprints so as to determine whether the current user incoming call voice data triggers an alarm or not.
The invention provides an offline detection method aiming at a mass offline voice file, and the offline voice file in a set time period is collected on time, and the offline voice file is preprocessed, batch voiceprint registered, batch voiceprint compared, speaker clustered and mined, so that potential illegal users can be further mined out, a specific list library is updated, and the accuracy of a subsequent real-time detection result is ensured.
According to the invention, through speaker clustering, speaker grouping is carried out on massive offline voice files, and specific list data mining is carried out according to a plurality of set mining strategies and combined with caller information, so that the accuracy of specific list mining is ensured.
Drawings
FIG. 1 is a schematic diagram of a system architecture of the present invention;
FIG. 2 is a flow chart of the method for detecting the operation of the illegal service on line in the invention;
FIG. 3 is a flow chart of an off-line detection method for illegal service operation in the present invention;
FIG. 4 is a schematic diagram of an application framework for online detection in the first embodiment;
FIG. 5 is a flow chart of an application architecture for online detection in the first embodiment;
FIG. 6 is a schematic diagram illustrating the operation of the on-line voiceprint unit in accordance with the first embodiment;
FIG. 7 is a schematic diagram showing a sound pretreatment process in the first embodiment;
FIG. 8 is a schematic diagram of a voiceprint feature extraction process in accordance with the first embodiment;
FIG. 9 is a schematic diagram of a voice print registration process in the first embodiment;
FIG. 10 is a flow chart of voice print real-time registration in the first embodiment;
FIG. 11 is a flow chart of the real-time voice print verification in the first embodiment;
FIG. 12 is a schematic diagram of an application framework for offline detection in the second embodiment;
FIG. 13 is a schematic diagram illustrating the operation of an off-line voiceprint unit in accordance with the second embodiment;
fig. 14 is a schematic diagram of a sound pretreatment process in the second embodiment;
FIG. 15 is a schematic diagram of a voice print registration process in the second embodiment;
FIG. 16 is a flow chart of offline registration of batch voiceprints in the second embodiment;
FIG. 17 is a flowchart of speaker clustering in the second embodiment;
fig. 18 is a schematic flow chart of specific list mining in the second embodiment;
the figure indicates: 1. the device comprises an application end, a voiceprint device, a recording acquisition device, an online voiceprint unit, an offline voiceprint unit and a voiceprint unit, wherein the voiceprint device comprises an application end, a voiceprint device, a recording acquisition device, an online voiceprint unit and an offline voiceprint unit.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples.
As shown in fig. 1, the system for detecting the operation of the illegal service comprises an application end 1 and a voiceprint device 2 which are connected with each other, wherein the application end 1 transmits voice data or offline voice files of a coming user to the voiceprint device 2, the voiceprint device 2 analyzes and processes the voice data or offline voice files of the coming user to output corresponding detection results, and the detection results and corresponding instructions are transmitted to the application end 1.
The application end 1 is provided with a recording acquisition device 11 for collecting voice data of an incoming call user, the voiceprint device 2 comprises an online voiceprint unit 21 and an offline voiceprint unit 22, the online voiceprint unit 21 is used for detecting online real-time illegal operations, and the offline voiceprint unit 22 is used for detecting offline massive voices.
Specifically, the online voiceprint unit 21 includes a preprocessing module, a voiceprint feature extraction module, a voiceprint registration module, a verification comparison module, and a database, where the preprocessing module is used to remove non-user sounds in the voice data of the incoming call user; the voiceprint feature extraction module is used for extracting voiceprint features of the incoming call user from the preprocessed voice data of the incoming call user; the voiceprint registration module is used for storing voiceprint features of the incoming call user in a database and establishing a unique mapping relation between the voiceprint features of the incoming call user and a user Identity (ID); the verification comparison module is used for verifying and comparing voiceprint features of the incoming call user with voiceprints corresponding to the specific list library and the user identity ID so as to output corresponding detection results; the database is internally provided with a specific list library and a user Identity (ID) -voiceprint library.
The offline voiceprint unit 22 comprises a preprocessing module, a batch voiceprint registration module, a batch voiceprint comparison module, a clustering module, an excavating module and a database, wherein the preprocessing module is used for removing non-user sounds in offline voice files; the batch voiceprint registration module is used for carrying out offline batch scanning, batch extraction and batch registration on the preprocessed offline voice files, and storing voiceprint characteristics corresponding to file names in a database; the batch voiceprint comparison module is used for carrying out pairwise combination comparison on voiceprints which are registered in batch, and outputting to obtain voiceprint comparison scores corresponding to a plurality of combinations; the clustering module outputs a caller grouping result by adopting a merging and classifying mode according to the comparison scores of the voiceprints; and the mining module outputs a specific list according to the grouping result of the talkers and the set mining strategy, and stores the specific list in the database.
The system is applied to practice to realize an on-line detection method for illegal service operation, as shown in fig. 2, and comprises the following steps:
a1, transmitting incoming user voice data to an online voiceprint unit in real time by an application end;
a2, after the online voiceprint unit receives the voice data of the incoming call user, voice preprocessing, voiceprint feature extraction, voiceprint registration and voiceprint verification comparison operation are sequentially carried out, and an online detection result of the illegal operation is obtained and transmitted to the application end.
The system is applied to practice to realize an off-line detection method for illegal service operation, as shown in fig. 3, and comprises the following steps:
b1, the application end transmits the collected offline voice file in the set time period to an offline voiceprint unit;
and B2, after the offline voiceprint unit receives the offline voice file, carrying out voice preprocessing, batch voiceprint registration, batch voiceprint comparison, speaker clustering and mining operation in sequence, and outputting to obtain specific list data.
The first embodiment and the second embodiment apply the above technical solutions, and, for a specific service scenario of a first specific person, based on a voiceprint recognition technology, perform real-time early warning or offline mining on a specific behavior of the first specific person from the following two aspects: the first embodiment is that the user voice and a specific list voiceprint library stored by a voiceprint system are compared in real time on line, and a result is returned in time; in the second embodiment, all voice files of all caller in a certain period are classified and compared according to speaker (i.e. talker) in offline mode, and then the potential fraudsters are further mined by combining with the business rules.
Example 1
As shown in fig. 4, the application system firstly transmits the voice from the user to a voiceprint device for mining a first specific person, the voiceprint device extracts the voice of the user, compares the voice with the voiceprint of the specific list stored in the device, and if the voice is not in the specific list, processes related services for the user according to the comparison result; if the service is in the specific list, the relevant service is not transacted for the service.
More specifically, the flow of detecting a first specific person using the voiceprint apparatus is shown in fig. 5. The voice stream is also sent to the TI text irrelevant voiceprint engine in the voiceprint device during the call with the seat, the voiceprint engine compares and detects the specific list stored in the user voiceprint system, if the specific list is matched with the voiceprint in the specific list library, the specific list alarm is triggered, otherwise, the normal voiceprint verification is carried out.
FIG. 6 shows the overall operation of the real-time detection voiceprint apparatus (i.e., the online voiceprint unit), which mainly includes the following steps:
1.1.1 Sound entry
The voice recording is a process of collecting voice of a user in real time by using professional recording equipment or a specially developed recording collection system in the process of communication between the user and the seat.
1.1.2 Sound pretreatment
The voice preprocessing refers to removing other voices of non-users. The specific preprocessing design is shown in fig. 7, wherein silence detection is to delete excessive silence, and voice quality detection is to analyze the signal-to-noise ratio of voice, and only the voice with higher signal-to-noise ratio meets the requirement of a voiceprint device on the voice quality; and speaker separation is to remove other voices than the user himself in the collected voice.
1.1.3 voiceprint feature extraction
As shown in fig. 8, voiceprint feature extraction refers to a process of extracting and modeling unique sound features associated with a user in sound using a deep neural network structure, for example, CNN, TDNN in combination with an attention mechanism, etc. The output of voiceprint feature extraction is the voiceprint corresponding to the speech, typically represented by a feature vector such as x-vector or d-vector. The extracted voiceprint features are closely related to the user, are biological features of the user, and have uniqueness.
1.1.4 voiceprint registration
Voiceprint registration means that the extracted voiceprint features of the user are stored in a database in a binary unreadable form, and a unique mapping relation is established between the voiceprint features and the user identity ID. After the voiceprint of the user is registered, the voiceprint is permanently stored in the database and can be accessed through the user identity ID, and the real-time voiceprint registration design is shown in FIG. 9.
1.1.4.1 real-time voiceprint registration process
The process of the user real-time voiceprint registration is completed by the aid of the seat in the process of the user and the seat. The computer operating system of the seat is provided with a page specially used for processing voiceprint operation, and the page is provided with function buttons related to voiceprint such as 'inquiry', 'registration', 'verification', and the like.
The real-time registration flow is shown in fig. 10, after the user accesses the customer service system, the seat first clicks "inquiry" to see whether the user has voiceprint. If the user has a voiceprint, the seat can trigger a "verify" operation request. If there is no voiceprint, the seat can trigger a "registration" request. The seat-triggered voiceprint operation request is sent to the voiceprint device to complete the relevant operation. In the registration link, the customer service firstly carries out manual check on the user, and after the user passes through the check on the user, a voiceprint registration request is triggered, so that in order to collect the voice with enough length, the customer service is recommended to ask questions of the user as much as possible to register. If the core does not pass, a cancel request is triggered, and the background system automatically stops or deletes the record just.
1.1.5 voiceprint verification
As shown in fig. 11, voiceprint verification refers to that during the conversation between the agent and the user, the system collects the voice of the user in real time in the background, and synchronously compares and verifies the voice with the voiceprints corresponding to the specific list library and the user ID. If the voice print is matched with the voice print in the specific list library, triggering specific list alarm to indicate that the malicious invasion of the first specific person is detected, otherwise, performing normal voice print verification and returning a corresponding verification result.
Example two
As shown in fig. 12, a mass of voices in a certain period of time are collected from the credit card call center in advance, and sent to a voiceprint device (i.e., an off-line voiceprint unit) for mining a first specific person, the voiceprint device analyzes and processes the mass of voices, and finally the mining result of the first specific person is output.
The specific working process of the voiceprint apparatus for offline mining of a first particular person is shown at 13 and mainly comprises:
2.1.1 offline Mass Voice File reading
The mass multi-user offline voice files are derived from daily mass recording files of a customer service center.
2.1.2 Sound File pretreatment
The voice preprocessing refers to removing other voices of non-users. The specific preprocessing design is shown in fig. 14, wherein silence detection is to delete excessive silence, and voice quality detection is to analyze the signal-to-noise ratio of voice, and only the voice with higher signal-to-noise ratio meets the requirement of a voiceprint device on the voice quality; and speaker separation is to remove other voices than the user himself in the collected voice.
2.1.3 offline batch voiceprint registration
And carrying out offline batch scanning on daily mass recording files of the customer service center, and carrying out voiceprint modeling on all voices. As shown in fig. 15, voiceprint registration refers to storing extracted voiceprint features of a user in a database in a binary non-readable form, and establishing a unique mapping relationship between the voiceprint features and file names.
2.1.3.1 offline voiceprint registration procedure
The offline registration process is shown in fig. 16, where the system first obtains information of voice files in batches, in particular file name information, caller number information, area information, caller time information, claimed user ID, etc., then extracts voiceprints for these files in batches, then performs batch registration, writes the voiceprint features of each file into the database, and maps the voiceprint features of each file with the corresponding voice file name.
2.1.4 batch voiceprint comparison
And carrying out pairwise comparison on the voiceprints which are registered in batches, and obtaining voiceprint comparison score results of all pairwise combinations.
2.1.5 talker clustering
The speaker clustering means that the voices belonging to the same speaker are continuously combined and classified by utilizing the voice print score results of the pairwise comparison of all voice files through a speaker clustering algorithm, finally, all voices belonging to the first speaker are classified into one class, all voices belonging to the second speaker are classified into one class, and the like, the similarity of voices in the class is extremely high, the similarity of voices between classes is low, and the purpose of grouping and arranging the voice fragments according to different speakers is achieved. The concrete speaker clustering flow is shown in fig. 17.
2.1.6 first specific personnel mining strategy
The first specific person mining strategy refers to that according to the obtained massive voice files and the speaker grouping result, the mining strategy shown in fig. 18 is used for judging, and finally the mined first specific person list is output.
The number policy refers to that the files grouped by each speaker are scanned for incoming call number information of all voice files in the group, and if the voice files in the group come from more than 2 different numbers, the speaker is listed in a suspected first specific person result list.
The "ID policy" refers to that, for each speaker, the user ID information of all voice files in the group is scanned, and if the voice files in the group come from more than 2 different user IDs, the speaker will be listed in the list of suspected first specific person results.
"frequency of call policy" refers to that, for each speaker group file, the incoming times of all voice files in the group are scanned, the incoming times of each voice file are statistically analyzed and converted into a frequency of calls (e.g., times/day), and for speakers with a frequency of calls above a certain threshold, a list of suspected first specific person results may be listed. The threshold value can be reasonably formulated according to specific business scenes and actual experience data.
The "area policy" refers to that the incoming call area information of all the voice files is obtained from the list result of the suspected first specific person, statistical analysis is performed on the area information, the duty ratio is calculated, if the duty ratio of a certain area exceeds a set threshold (the threshold can be specially formulated according to the actual business), the area can be listed as a risk area, and the result is output for reference when the subsequent customer service processes the incoming call, and the incoming call of the area is particularly focused on warning.
In summary, the present technical solution proposes a voiceprint modeling method, and uses a voiceprint technology to perform voiceprint modeling and warehousing on an caller in a credit card customer service center; providing a voice print-based speaker clustering algorithm, and grouping voices of callers in a credit card customer service center according to different speakers; providing a fraudster mining algorithm, analyzing and judging a grouping result of a speaker clustering algorithm, and finding out a fraudster; a mechanism of a voiceprint specific list is provided, and by creating the voiceprint specific list, the voice of a caller in a credit card customer service center is compared with a specific list library in real time, and a fraudster is warned in real time. Therefore, the offline mass voice can be subjected to speaker clustering to realize the mining of the first specific person, the specific business of the first specific person is early warned in real time by utilizing a specific list mechanism, and the mining of the first specific person is realized by combining caller information and a voiceprint speaker clustering method. The method provides a complete and feasible technical scheme for solving illegal personnel illegal business operation, and is a qualitative change from 0 to 1 and a complete innovation in technology compared with the prior art.
Claims (10)
1. The illegal service operation detection system is characterized by comprising an application end (1) and a voiceprint device (2) which are connected with each other, wherein the application end (1) transmits voice data or offline voice files of a calling user to the voiceprint device (2), and the voiceprint device (2) analyzes and processes the voice data or offline voice files of the calling user so as to output corresponding detection results and transmits the detection results and corresponding instructions to the application end (1).
2. The system for detecting the operation of the illegal service according to claim 1, wherein the application end (1) is provided with a recording acquisition device (11) for collecting voice data of an incoming user.
3. The system for detecting the operation of a violation business according to claim 1, wherein the voiceprint device (2) comprises an online voiceprint unit (21) and an offline voiceprint unit (22), the online voiceprint unit (21) is used for detecting the operation of a violation on line in real time, and the offline voiceprint unit (22) is used for detecting the operation of a violation on offline massive voice.
4. A system for detecting a violation business operation according to claim 3, characterized in that the online voiceprint unit (21) comprises a preprocessing module, a voiceprint feature extraction module, a voiceprint registration module, a verification comparison module and a database, the preprocessing module being configured to remove non-user sounds in incoming user speech data; the voiceprint feature extraction module is used for extracting voiceprint features of the caller from the preprocessed caller voice data; the voiceprint registration module is used for storing voiceprint features of the incoming call user in a database and establishing a unique mapping relation between the voiceprint features of the incoming call user and a user Identity (ID); the verification comparison module is used for verifying and comparing the voice print characteristics of the incoming call user with voice prints corresponding to the specific list library and the user identity ID so as to output corresponding detection results; and a specific list library and a user identity ID-voiceprint library are arranged in the database.
5. A system for detecting a violation business operation according to claim 3, characterized in that the offline voiceprint unit (22) comprises a preprocessing module, a batch voiceprint registration module, a batch voiceprint comparison module, a clustering module, an excavating module and a database, the preprocessing module being configured to remove non-user sounds in an offline voice file; the batch voiceprint registration module is used for carrying out offline batch scanning, batch extraction and batch registration on the preprocessed offline voice files, and storing voiceprint characteristics corresponding to file names in a database; the batch voiceprint comparison module is used for carrying out pairwise combination comparison on voiceprints which are registered in batch, and outputting to obtain voiceprint comparison scores corresponding to a plurality of combinations; the clustering module outputs a caller grouping result by adopting a merging and classifying mode according to the voiceprint comparison scores; and the mining module outputs a specific list according to the grouping result of the talkers and the set mining strategy, and stores the specific list in the database.
6. An on-line detection method for illegal service operation is characterized by comprising the following steps:
a1, transmitting incoming user voice data to an online voiceprint unit in real time by an application end;
a2, after the online voiceprint unit receives the voice data of the incoming call user, voice preprocessing, voiceprint feature extraction, voiceprint registration and voiceprint verification comparison operation are sequentially carried out, and an online detection result of the illegal operation is obtained and transmitted to the application end.
7. The method for online detection of operation of illegal services according to claim 6, wherein said step A2 specifically comprises the steps of:
a21, preprocessing the voice data of the incoming call user to remove non-user voice, wherein the preprocessing process comprises silence detection, voice quality detection, noise removal and speaker separation;
a22, extracting voice print characteristics of the incoming call user from the preprocessed voice data of the incoming call user based on the deep neural network;
a23, storing the extracted voiceprint features of the incoming call user into a database in a binary unreadable form, and establishing a unique mapping relation between the voiceprint features of the incoming call user and the user identity ID so as to finish online voiceprint registration;
a24, comparing the voiceprint features of the incoming call user with the voiceprint features in the specific list library, and outputting alarm information to the application end if the voiceprint features of the incoming call user are judged to belong to the specific list;
otherwise, comparing the voiceprint features of the incoming call user with the voiceprint features of the corresponding user Identity (ID) in the database, and outputting a corresponding comparison result to the application end.
8. The off-line detection method for the operation of the illegal service is characterized by comprising the following steps of:
b1, the application end transmits the collected offline voice file in the set time period to an offline voiceprint unit;
and B2, after the offline voiceprint unit receives the offline voice file, carrying out voice preprocessing, batch voiceprint registration, batch voiceprint comparison, speaker clustering and mining operation in sequence, and outputting to obtain specific list data.
9. The method for off-line detection of operation of a violation business according to claim 8, wherein the step B2 specifically comprises the steps of:
b21, preprocessing the offline voice file to remove non-user sounds, wherein the preprocessing process comprises silence detection, voice quality detection, noise removal and speaker separation;
b22, carrying out batch scanning and batch extraction on the preprocessed offline voice file, then storing the extracted voice characteristic into a database in a binary unreadable form, and establishing a unique mapping relation between the voice characteristic and a corresponding offline voice file name to sequentially finish offline batch voice registration;
b23, carrying out pairwise combination comparison on offline voiceprints which are registered in batches to obtain a plurality of combination voiceprint comparison scores;
b24, according to a plurality of combined voiceprint comparison score results, combining and classifying voices belonging to different speakers, and outputting to obtain speaker grouping results;
and B25, outputting and obtaining specific list data according to a caller grouping result and a set mining strategy, wherein the mining strategy comprises a number strategy, an ID strategy, a calling frequency strategy and an area strategy.
10. The method for offline detection of operation of illegal services according to claim 9, wherein the number policy is specifically: for each caller group file, scanning caller number information of all voice files in the group, if the voice files in the group come from more than 2 different numbers, putting the caller into a suspected specific list;
the ID strategy specifically comprises the following steps: for each speaker group file, scanning user ID information of all voice files in the group, and if the voice files in the group come from more than 2 different user IDs, putting the speaker into a suspected specific list;
the calling frequency strategy specifically comprises the following steps: aiming at the files grouped by each caller, scanning the incoming call time of all voice files in the group, carrying out statistical analysis on the incoming call time of each voice file, converting the incoming call time into calling frequency, and listing the callers corresponding to the calling frequency higher than the set calling frequency threshold value in a suspected specific list;
the regional policy specifically includes: and acquiring incoming call region information of all the voice files from the suspected specific list result, carrying out statistical analysis on the region information, calculating the duty ratio, and if the duty ratio of a certain region exceeds a set region duty ratio threshold value, listing the region as a risk region and outputting the result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211099629.7A CN116665708A (en) | 2022-09-07 | 2022-09-07 | Illegal service operation detection system and method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211099629.7A CN116665708A (en) | 2022-09-07 | 2022-09-07 | Illegal service operation detection system and method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116665708A true CN116665708A (en) | 2023-08-29 |
Family
ID=87722974
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211099629.7A Pending CN116665708A (en) | 2022-09-07 | 2022-09-07 | Illegal service operation detection system and method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116665708A (en) |
-
2022
- 2022-09-07 CN CN202211099629.7A patent/CN116665708A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107222865B (en) | Communication swindle real-time detection method and system based on suspicious actions identification | |
| US9113001B2 (en) | Systems, methods, and media for disambiguating call data to determine fraud | |
| US7386105B2 (en) | Method and apparatus for fraud detection | |
| US9503571B2 (en) | Systems, methods, and media for determining fraud patterns and creating fraud behavioral models | |
| CN106550155B (en) | Swindle sample is carried out to suspicious number and screens the method and system sorted out and intercepted | |
| US9240013B2 (en) | Evaluation of voice communications | |
| CN109036435B (en) | Identity authentication and identification method based on voiceprint information | |
| US8793131B2 (en) | Systems, methods, and media for determining fraud patterns and creating fraud behavioral models | |
| US8078463B2 (en) | Method and apparatus for speaker spotting | |
| US20150381801A1 (en) | Systems, methods, and media for disambiguating call data to determine fraud | |
| CN109509093B (en) | Transaction security control method and system based on main body portrait | |
| CN110533288A (en) | Business handling process detection method, device, computer equipment and storage medium | |
| CN107197463A (en) | A kind of detection method of telephone fraud, storage medium and electronic equipment | |
| CN110248322A (en) | A kind of swindling gang identifying system and recognition methods based on fraud text message | |
| CN111833068A (en) | Identity verification system and method based on voiceprint recognition | |
| CN114971627A (en) | Data monitoring system and method based on computer network | |
| CN117391723A (en) | Anti-fraud platform and method | |
| CN115103063B (en) | Internet fraud identification method and device for fake customer service class | |
| CN112925971A (en) | Multi-source analysis-based fraud topic detection method and device | |
| CN116665708A (en) | Illegal service operation detection system and method thereof | |
| CN116963072A (en) | Fraud user early warning method and device, electronic equipment and storage medium | |
| CN111161743B (en) | Cash receiving supervision method and system based on voice recognition | |
| CN113763963B (en) | Phone card fraud detection method, system, computer equipment and storage medium | |
| CN119832934A (en) | Account anomaly determination method and device, storage medium and electronic equipment | |
| US12323554B1 (en) | Audio speech signal analysis for fraud detection |
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 |