CN111212193B - Method for identifying user on-hook state based on user color ring information - Google Patents
Method for identifying user on-hook state based on user color ring information Download PDFInfo
- Publication number
- CN111212193B CN111212193B CN202010037240.4A CN202010037240A CN111212193B CN 111212193 B CN111212193 B CN 111212193B CN 202010037240 A CN202010037240 A CN 202010037240A CN 111212193 B CN111212193 B CN 111212193B
- Authority
- CN
- China
- Prior art keywords
- user
- hook
- hook state
- call center
- information
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M19/00—Current supply arrangements for telephone systems
- H04M19/02—Current supply arrangements for telephone systems providing ringing current or supervisory tones, e.g. dialling tone or busy tone
- H04M19/04—Current supply arrangements for telephone systems providing ringing current or supervisory tones, e.g. dialling tone or busy tone the ringing-current being generated at the substations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/42—Systems providing special services or facilities to subscribers
- H04M3/42017—Customized ring-back tones
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Telephonic Communication Services (AREA)
Abstract
The invention obtains the user state through the color ring in the media, can ensure that the call center can obtain the reliable on-hook state of the user, can provide reliable data in the scenes of automatic outbound, AI outbound and the like after obtaining the correct on-hook state, can normally carry out continuous statistics and analysis, simultaneously saves a large amount of manpower for carrying out manual marking, can return error codes only after the color ring is played, can obtain the on-hook state of the user in advance by identifying the color ring, and also improves the efficiency of the system.
Description
Technical Field
The invention belongs to the technical field of communication, and particularly relates to a method for identifying a user on-hook state based on user polyphonic ringtone information.
Background
In the existing call, the call is divided into 2 parts, one part is signaling, and the other part is media; the signaling is used for controlling a call flow and comprises the steps of call origination, media negotiation, connection, hang-up and the like; the media is used to establish media transmission channel for media transmission, the media includes 2 types, color ring and call, in normal call, the telephone user can judge the state of the called user is busy or off or other according to the color ring. The call center generally uses the hang-up code in the signaling to judge the hang-up state of the called subscriber.
The existing problem is that the hang-up code in the signaling is often inconsistent with the real hang-up reason, namely, the hang-up code is inaccurate, and the reason of the problem is mainly that equipment provided by a plurality of manufacturers exists in the core network of an operator, but the manufacturers directly have no uniform standard for defining the error code. Eventually leading to inconsistent error codes in the signaling and the true reason for hang-up.
The direct problem caused by inaccurate hang-up code is that the call center can not obtain the real state of the called user, and the later statistics and analysis results can be seriously influenced in the scenes of AI outbound and automatic outbound.
Disclosure of Invention
Technical problem to be solved
The invention provides a method for identifying a user on-hook state based on user color ring back tone information, which aims to solve the practical problems mentioned in the background technology.
(II) technical scheme
In order to achieve the purpose, the invention provides the following technical scheme: a method for identifying the on-hook state of a user based on the color ring information of the user comprises the following steps:
s1: collecting a large number of audio files of ring back tones of a user on-hook, such as empty numbers, shutdown, refusal of the user, temporary unavailable connection, wrong numbers and the like, and marking corresponding reasons for the user on-hook;
s2: collecting a large number of ring back tones of non-user on-hook, and carrying out non-user on-hook ring back tone marking;
s3: carrying out secondary processing on the original audio files collected in S1 and S2;
s4: pre-emphasis, framing, Hamming window adding, FFT, Mel filter bank processing, logarithm operation, DTF to extract the characteristics of the audio signal;
s5: constructing a neural network consisting of 3 full-connect layers, 1 bidirectional RNN containing 4 LSTM layers, 2 full-connect layers and a SoftMax layer;
s6: inputting the characteristic data obtained by the S4 into the neural network constructed by the S5 for training, and obtaining a user on-hook state classification model for classifying the polyphonic ringtone information;
s7: initiating an automatic outbound call at a call center, carrying out media negotiation when receiving 180/183 media information with SDP, establishing a voice channel, and sending the voice information to a user on-hook state classification model for processing after arranging the voice information;
s8: after the user on-hook state classification model identifies the user on-hook state, the call center is informed to carry out corresponding operation;
s9: after receiving the 200 response messages, the call center informs the user of the on-hook state classification model and terminates the identification;
s10: and after receiving the BYE signaling, the call center informs the user of the on-hook state classification model and terminates the identification.
Further, the step S1 specifically includes: calling out blank numbers in batches through a call center, shutting down, stopping, refusing the user, temporarily failing to connect, and storing the polyphonic ringtone audio called by the user through the call center, and automatically marking.
Further, step S2 specifically includes: and the call center calls out in batch, records the color ring of all users, stores all the color rings receiving the 200 response messages, marks the color rings as off-hook, and discards other color rings.
Further, step S3 specifically includes: cutting off the mute at the tail end; and cutting the installation time of the audio file, wherein the time length is 7 seconds, the audio file is overlapped forwards for 2 seconds, and the original file label is inherited.
Further, the step S7 specifically includes the following steps:
step (1): decoding the voice information into S16 format in real time;
step (2): performing real-time audio signal feature extraction on the decoded rear-end audio;
and (3): and sending the extracted complete audio features of the next frame to the user on-hook state classification model.
Further, the step S8 specifically includes continuing to perform the next round of identification without notifying the call center when the user on-hook state classification model identifies that the user on-hook state is the non-on-hook state; and after the user on-hook state classification model shows that the user on-hook state is the on-hook state, the corresponding on-hook reason is sent to the call center, and the call center carries out corresponding on-hook operation and records the on-hook reason.
(III) compared with the prior art, has the following beneficial effects
(1) Compared with the method for judging the on-hook reason of the user through the signaling, the accuracy can be greatly improved by identifying the on-hook state through the polyphonic ringtone information of the user.
(2) Compared with the existing color ring identification technology on the market at present, the method for identifying the on-hook state of the color ring information of the user has the characteristics of high identification speed and strong processing performance, and has stable system and high cost performance.
(3) The method for identifying the on-hook state by the user color ring information can be combined with a forecast outbound system, the system can quickly identify the on-hook state of the user, and when the user hangs up, the system automatically jumps to the next number to continue the outbound. Therefore, in the process of predicting the outbound call, the on-hook state of the user can be rapidly and accurately identified, and the effective call completing rate of the customer service can be greatly improved.
(4) The method uses the same set of bottom layer codes with the call center, has the system stability of telecommunication level, does not need a large amount of hardware equipment, and has good applicability, expandability and high cost performance.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a diagram of the steps of carrying out the present invention;
FIG. 3 is a diagram of a neural network of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the present invention provides a technical solution: a method for identifying the on-hook state of a user based on color ring back tone information of the user comprises the following specific steps:
s1, collecting a large number of blank numbers, shutting down, stopping, refusing the user, temporarily not being communicated, wrong numbers and other audio files of the polyphonic ringtone of the user, and marking the corresponding reason for hanging up the user;
step S1 specifically includes: calling out blank numbers in batch through a call center, shutting down, stopping, refusing the user, temporarily failing to connect, and automatically marking the telephone number with wrong number, and then storing the frequency of the user calling the color ring tone through the call center.
S2, collecting a large amount of color ring back tones of non-user on-hook, and labeling the color ring back tones of the non-user on-hook;
step S2 specifically includes:
s2-1, calling out in batch by the call center, recording all user color ring, storing all color ring which receives 200 response message, marking as off-hook, and discarding other color ring.
S3, carrying out secondary processing on the original audio files collected in S1 and S2;
step S3 specifically includes:
s3-1, cutting off the head end and the tail end by muting
Calculating the energy of each frame from head to tail, and deleting all frames smaller than the threshold;
and S3-2, cutting the installation time of the audio file, wherein the installation time is 7 seconds, the installation time is overlapped for 2 seconds, and the original file label is inherited.
S4, pre-emphasis, framing, Hamming window addition, FFT, Mel filter bank processing, logarithm operation, DTF to extract the characteristics of the audio signal.
S4-1, passing the voice signal through the transfer function as H (z) -1-muz-1The first order high pass filter of (1) performing pre-emphasis, wherein μ ═ 0.95;
s4-2, framing the audio signal by 256 sampling points, namely, the time of each frame is 32ms, the formula is t-N/f, and N is the number of the sampling points;
s4-3, multiplying each frame by a Hamming window to increase the continuity of the frame; the result is that
N is the number of sampling points, and a is 0.5;
s4-4, by
Performing Fourier transform;
s4-5, smoothing the frequency spectrum through a Mel frequency filter bank, eliminating harmonic waves, highlighting the formants of the original voice, and defining the frequency response as:
where m is the number of filters, f (m) floor ((nfft +1) × (i)/samplerate, nfft is the number of ffts, h (i) is the m +2 average of the Mel scale values for the minimum and maximum frequencies, samplerate is the sampling rate;
s4-6, calculating logarithmic energy for each filter bank output
s4-7, performing DCT to obtain coefficients:
s5, constructing a neural network consisting of 3 full-connect layers, 1 bidirectional RNN containing 4 LSTM layers, 2 full-connect layers and a SoftMax layer; a Long Short-Term Memory network (LSTM), a time-cycled neural network;
s5-1, establishing full-connect layers of 3 1024 nodes, and performing Batch Normalization operation behind each full-connect layer to accelerate training and improve accuracy, referring to the full-connect layer in FIG. 3;
s5-2, then establishing a bidirectional RNN containing 4 layers of LSTM, wherein the number of Cells is 1024, and the RNN layer in the graph of FIG. 3 is referred to;
s5-3, establishing full-connect layers of 2 1024 nodes, and performing Batch Normalization operation behind each full-connect layer to accelerate training and improve accuracy;
s5-4, finally adding a SoftMax layer, and defining a loss function as
S6, inputting the characteristic data obtained through S4 into the neural network constructed in S5 for training, and obtaining a user on-hook state classification model for classifying the polyphonic ringtone information;
s7, initiating automatic outbound at the call center, carrying out media negotiation when receiving 180/183 media information with SDP, establishing a voice channel, and sending the voice information to a user on-hook state classification model for processing after being sorted;
step S7 specifically includes:
s7-1, decoding the voice information into S16 format in real time;
s7-2, performing real-time audio signal feature extraction on the decoded rear-end audio;
and S7-3, extracting the complete audio feature of the next frame and then sending the extracted complete audio feature to the user on-hook state classification model.
S8, after the user on-hook state is identified by the user on-hook state classification model, informing the call center to perform corresponding operation;
step S8 specifically includes:
s8-1, after the user on-hook state classification model identifies that the user on-hook state is a non-on-hook state, continuing to perform the next round of identification without informing the call center;
and S8-2, after the user on-hook state classification model shows that the user on-hook state is the on-hook state, sending the corresponding on-hook reason to the call center, and the call center performs the corresponding on-hook operation and records the on-hook reason.
And S9, after receiving the 200 response message, the call center informs the user of the on-hook state classification model and terminates the identification.
And S10, after receiving the BYE signaling, the call center informs the user of the on-hook state classification model and terminates the identification.
The method comprises the steps of collecting a large number of blank numbers, shutting down, stopping, refusing user connection, temporarily failing to connect, wrong number and other audio files of the polyphonic ringtone hung up by the user, carrying out corresponding reason marking of the hung up user, collecting a large number of polyphonic ringtones hung up by the non-user, and carrying out polyphonic ringtone marking of the hung up polyphonic ringtone not by the user; pre-emphasis, framing, Hamming window adding, FFT, Mel filter bank processing, logarithm operation, DTF to extract the characteristics of the audio signal.
The invention initiates automatic outbound at the call center, when receiving 180/183 media information with SDP, carries out media negotiation, establishes a voice channel, arranges the voice information and sends the voice information to a user on-hook state classification model for processing; after the user on-hook state classification model identifies the user on-hook state, the call center is informed to carry out corresponding operation; after receiving the 200 response messages, the call center informs the user of the on-hook state classification model and terminates the identification; and after receiving the BYE signaling, the call center informs the user of the on-hook state classification model and terminates the identification.
In the embodiment, 10000 test numbers are tested by identifying the on-hook state of the user based on the polyphonic ringtone information, and the identification accuracy reaches 99%; the method for identifying the on-hook state of the user based on the color ring information of the user is used in the process of predicting the outbound, and the proper outbound prediction rule is automatically calculated by combining the characteristics of variable domestic services and multiple data uncertainty factors, so that the balance of the outbound speed, the seat workload and the customer experience degree is realized to the maximum extent.
In the description herein, references to the description of "one embodiment," "an example," "a specific example" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims (6)
1. A method for identifying the on-hook state of a user based on the color ring information of the user is characterized by comprising the following steps:
s1: collecting an empty number, shutting down, stopping, refusing the user, temporarily not connecting, and marking a corresponding user on-hook reason, wherein the number of the audio file of the polyphonic ringtone of the user with the wrong number is wrong;
s2: collecting polyphonic ringtone of a non-user on-hook, and marking the polyphonic ringtone of the non-user on-hook;
s3: carrying out secondary processing on the original audio files collected in S1 and S2;
s4: pre-emphasis, framing, Hamming window adding, FFT, Mel filter bank processing, logarithm operation, DTF to carry out audio signal feature extraction;
s5: constructing a neural network consisting of 3 full-connect layers, 1 bidirectional RNN containing 4 LSTM layers, 2 full-connect layers and a SoftMax layer;
s6: inputting the characteristic data obtained by the S4 into the neural network constructed by the S5 for training, and obtaining a user on-hook state classification model for classifying the polyphonic ringtone information;
s7: initiating an automatic outbound call at a call center, carrying out media negotiation when receiving 180/183 media information with SDP, establishing a voice channel, and sending the voice information to a user on-hook state classification model for processing after arranging the voice information;
s8: after the user on-hook state classification model identifies the user on-hook state, the call center is informed to carry out corresponding operation;
s9: after receiving the 200 response messages, the call center informs the user of the on-hook state classification model and terminates the identification;
s10: and after receiving the BYE signaling, the call center informs the user of the on-hook state classification model and terminates the identification.
2. The method for identifying the on-hook status of the user based on the polyphonic ringtone information of the user according to claim 1, wherein the step S1 specifically includes: calling out blank numbers in batches through a call center, shutting down, stopping, refusing the user, temporarily failing to connect, and storing the polyphonic ringtone audio called by the user through the call center, and automatically marking.
3. The method for identifying the on-hook status of the user based on the polyphonic ringtone information of the user according to claim 1, wherein the step S2 specifically comprises: and the call center calls out in batch, records the color ring of all users, stores all the color rings receiving the 200 response messages, marks the color rings as off-hook, and discards other color rings.
4. The method for identifying the on-hook status of the user based on the polyphonic ringtone information of the user according to claim 1, wherein the step S3 specifically comprises: cutting off the mute at the tail end; and cutting the installation time of the audio file, wherein the time length is 7 seconds, the audio file is overlapped forwards for 2 seconds, and the original file label is inherited.
5. The method for identifying the on-hook status of the user based on the polyphonic ringtone information of the user according to claim 1, wherein the step S7 specifically comprises the following steps:
step (1): decoding the voice information into S16 format in real time;
step (2): performing real-time audio signal feature extraction on the decoded rear-end audio;
and (3): and sending the extracted complete audio features of the next frame to the user on-hook state classification model.
6. The method of claim 1, wherein the method for identifying the on-hook status of the user based on the color ring back tone information of the user comprises: the step S8 specifically includes continuing the next round of recognition without notifying the call center after the user on-hook state classification model recognizes that the user on-hook state is the non-on-hook state; and after the user on-hook state classification model shows that the user on-hook state is the on-hook state, the corresponding on-hook reason is sent to the call center, and the call center carries out corresponding on-hook operation and records the on-hook reason.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010037240.4A CN111212193B (en) | 2020-01-14 | 2020-01-14 | Method for identifying user on-hook state based on user color ring information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010037240.4A CN111212193B (en) | 2020-01-14 | 2020-01-14 | Method for identifying user on-hook state based on user color ring information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111212193A CN111212193A (en) | 2020-05-29 |
| CN111212193B true CN111212193B (en) | 2021-04-27 |
Family
ID=70789047
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010037240.4A Active CN111212193B (en) | 2020-01-14 | 2020-01-14 | Method for identifying user on-hook state based on user color ring information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111212193B (en) |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1400797A (en) * | 2001-08-06 | 2003-03-05 | 赵宁 | Device for treating calling information of telephone |
| CN100486274C (en) * | 2002-10-28 | 2009-05-06 | 中兴通讯股份有限公司 | Method of realizing multiside calling business in flexible exchange system |
| JP2007142845A (en) * | 2005-11-18 | 2007-06-07 | Matsushita Electric Ind Co Ltd | Exchange device and its control method |
| CN101702831A (en) * | 2009-09-10 | 2010-05-05 | 厦门敏讯信息技术股份有限公司 | Wireless equipment and call control method thereof |
| CN104093213A (en) * | 2014-08-04 | 2014-10-08 | 耿直 | Dispatching communication method and public network dispatching communication system based on public network transmission |
| CN106453927A (en) * | 2016-10-25 | 2017-02-22 | 广东亿迅科技有限公司 | A number state automatic identification method and apparatus based on a PSTN network |
| CN109688575B (en) * | 2018-11-12 | 2022-10-25 | 平安科技（深圳）有限公司 | Number state identification method and device, terminal equipment and storage medium thereof |
| CN110290280B (en) * | 2019-05-28 | 2021-08-13 | 同盾控股有限公司 | Terminal state identification method and device and storage medium |
-
2020
- 2020-01-14 CN CN202010037240.4A patent/CN111212193B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111212193A (en) | 2020-05-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105979106B (en) | A kind of the ringing tone recognition methods and system of call center system | |
| CN104936182B (en) | A kind of method and system of intelligence management and control fraudulent call | |
| CN107331385A (en) | A kind of identification of harassing call and hold-up interception method | |
| CN102404462B (en) | Call progress analyzing method for phone dialing system and device | |
| US8391445B2 (en) | Caller identification using voice recognition | |
| CN106936997B (en) | A kind of rubbish voice recognition methods and system based on social networks map | |
| CN110072019A (en) | A kind of method and device shielding harassing call | |
| CN100496070C (en) | Identification system of automatic answering equipment and its identification method | |
| CN104410973B (en) | A kind of fraudulent call recognition methods of playback and system | |
| EP0663783B1 (en) | Disconnect signalling detection arrangement | |
| US7190768B2 (en) | Method and apparatus for human-assisted adaptive outbound telephone call classification based on pattern recognition | |
| CN111601002B (en) | Client number state matching method and device, electronic equipment and storage medium | |
| US5949864A (en) | Fraud prevention apparatus and method for performing policing functions for telephone services | |
| CN106453927A (en) | A number state automatic identification method and apparatus based on a PSTN network | |
| CN111212193B (en) | Method for identifying user on-hook state based on user color ring information | |
| CN111508527A (en) | Telephone answering state detection method, device and server | |
| CN114169438A (en) | Telecommunication network fraud identification method, device, equipment and storage medium | |
| CN111901488B (en) | Method for improving outbound efficiency of voice robot based on number state | |
| CN109756635A (en) | Phone customer service analogy method, device, storage medium and electronic equipment based on WEB | |
| CN113079259A (en) | Outbound failure result detection method based on freeswitch and ASR technology | |
| CN112511696A (en) | System and method for identifying bad content of call center AI engine | |
| CN110166637B (en) | Empty number identification method and device | |
| CN113438368B (en) | Method, device and computer readable storage medium for realizing ring back tone detection | |
| CN107370865A (en) | Recognition methods, device and the terminal of harassing call | |
| CN115659217A (en) | Fraud recognition model training method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |