CN109767335A - Double record quality detecting methods, device, computer equipment and storage medium - Google Patents
Double record quality detecting methods, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109767335A CN109767335A CN201811537633.0A CN201811537633A CN109767335A CN 109767335 A CN109767335 A CN 109767335A CN 201811537633 A CN201811537633 A CN 201811537633A CN 109767335 A CN109767335 A CN 109767335A
- Authority
- CN
- China
- Prior art keywords
- answer
- rule
- preset
- client
- emotional state
- 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
Abstract
The invention discloses a kind of double record quality detecting methods, device, computer equipment and storage mediums.The described method includes: sending the double record instructions of starting to client if receiving the financial product purchase request of client transmission;It is chosen from preset problem base and closes rule problem, and send client for voice the problem of closing rule problem by way of machine talk;The video image that client is sent inputs preset human face recognition model and carries out recognition of face;If face recognition result is that the identity of user is legal, preset speech recognition modeling is used, text conversion is carried out to the answer voice that client is sent, obtains the corresponding answer text of answer voice;Answer text is matched with rule answer is closed;If matching result is successful match, confirm that conjunction rule quality inspection passes through.Technical solution of the present invention realizes self-service double records and rule quality inspection is closed in automatic progress in real time, does not need manpower intervention, human cost is effectively reduced, and improve the efficiency closed and advise quality inspection.
Description
Technical field
The present invention relates to field of computer technology, more particularly to one kind pair to record quality detecting method, device, computer equipment and deposit
Storage media.
Background technique
Require to carry out insurance sales process double records, i.e. video and audio recording currently, silver protects prison, double records of insurance mainly for
The sales section of insurance products refers to that related marketing organization will acquire audiovisuals, electronics by technological means such as video and audio recordings
The mode of data, record and preservation insurance sales process key link, to realize that sales behavior can play back, important information can be looked into
The phenomenon that inquiry, problem responsibility can confirm, avoid irregularity generation.
With the continuous development of network application, the mode of online sale financial product is widely used, when passing through internet
When selling financial product, double record processes need to act on behalf of on-line finance complete by way of customer service casting problem user answers a question
At, and in the subsequent conjunction rule quality inspection by the double record videos completions of manual type access to double record contents, cause human cost higher
And close rule quality inspection inefficiency.
Summary of the invention
The embodiment of the present invention provides a kind of pair of record quality detecting method, device, computer equipment and storage medium, current to solve
Double records are carried out using manual type and advise quality inspection with artificial close is carried out afterwards, lead to human cost height and close rule quality inspection low efficiency
Problem.
A kind of double record quality detecting methods, comprising:
If receiving the financial product purchase request of client transmission, the double record instructions of starting are sent to the client,
Wherein, the double record instructions of the starting make the client open cam device and microphone apparatus;
If receiving double record start completion message that the client returns, according to preset selection mode, from presetting
The problem of library in choose and close rule problem, and described in by way of machine talk being sent voice the problem of the conjunctions rule problem to
Client;
Receive the answer voice and video image that the client is sent, wherein the vocal answer is the client
The voice for closing rule problem is answered by the user that the microphone apparatus acquires, the video image is that the client is logical
The user for crossing the cam device shooting answers the video for closing rule problem;
The video image is inputted into preset human face recognition model and carries out recognition of face, obtains face recognition result;
If the face recognition result is that the identity of the user is legal, preset speech recognition modeling is used, to institute
It states answer voice and carries out text conversion, obtain the corresponding answer text of the answer voice;
Obtain the conjunction rule problem corresponding conjunction rule answer from the preset problem base, and by the answer text with
The conjunction rule answer is matched, and matching result is obtained;
If the matching result is successful match, confirm that conjunction rule quality inspection passes through, and double to client transmission end
The instruction of record.
A kind of double record quality inspection devices, comprising:
Starting module, if the financial product purchase request for receiving client transmission, sends to the client
The double record instructions of starting, wherein the double record instructions of starting make the client open cam device and microphone apparatus;
Module is chosen, if the double record start completion message returned for receiving the client, according to preset choosing
Take mode, from preset problem base choose close rule problem, and by way of machine talk by the conjunction rule problem the problem of
Voice is sent to the client;
Receiving module, the answer voice and video image sent for receiving the client, wherein the vocal answer
It is the voice that the client answers the conjunction rule problem by the user that the microphone apparatus acquires, the video image is
The client answers the video for closing rule problem by the user that the cam device is shot;
Identification module carries out recognition of face for the video image to be inputted preset human face recognition model, obtains people
Face recognition result;
Conversion module uses preset voice if the identity for the face recognition result to be the user is legal
Identification model carries out text conversion to the answer voice, obtains the corresponding answer text of the answer voice;
Matching module, for obtaining the corresponding conjunction rule answer of the conjunction rule problem from the preset problem base, and will
The answer text is matched with conjunction rule answer, obtains matching result;
Module is completed in quality inspection, if being successful match for the matching result, confirms that conjunction rule quality inspection passes through, and to described
Client sends the instruction for terminating double records.
A kind of computer equipment, including memory, processor and storage are in the memory and can be in the processing
The computer program run on device, the processor realize the step of above-mentioned double record quality detecting methods when executing the computer program
Suddenly.
A kind of computer readable storage medium, the computer-readable recording medium storage have computer program, the meter
Calculation machine program realizes the step of above-mentioned double record quality detecting methods when being executed by processor.
In above-mentioned double record quality detecting methods, device, computer equipment and storage medium, the sale of financial product in internet
Cheng Zhong, if server-side receives the financial product purchase request of client transmission, to client send the double record instructions of starting with
The double records of starting, if the double record start completion message for receiving client return are asked according to preset selection mode from preset
It is chosen in exam pool and closes rule problem, send client for voice the problem of closing rule problem by way of machine talk, and receive
The answer voice for user's pairing rule problem that client returns and user answer video image when closing rule problem, and server-side is logical
It crosses human face recognition model and speech recognition modeling is passed through after confirming that the identity of user is legal to video image progress recognition of face
It identifies the corresponding answer text of answer voice of user, and answer progress is advised into the conjunction corresponding with rule problem is closed of answer text
Match, confirms that conjunction rule quality inspection passes through if successful match.It realizes and is automatically performed double record processes by way of self-service double records, and is right
The content of double records carries out closing rule quality inspection in real time automatically, so that double records and conjunction rule quality inspection do not need manpower intervention, effectively
Human cost is reduced, and improves the efficiency for closing rule quality inspection.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is an application environment schematic diagram of double record quality detecting methods in one embodiment of the invention;
Fig. 2 is a flow chart of double record quality detecting methods in one embodiment of the invention;
Fig. 3 is a flow chart of step S6 in double record quality detecting methods in one embodiment of the invention;
Fig. 4 is a flow chart of the step S7 of double record quality detecting methods in one embodiment of the invention;
Fig. 5 is a flow chart of step S71 in double record quality detecting methods in one embodiment of the invention;
Fig. 6 is a schematic diagram of double record quality inspection devices in one embodiment of the invention;
Fig. 7 is a schematic diagram of computer equipment in one embodiment of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
Double record quality detecting methods provided by the present application, can be applicable in application environment as shown in Figure 1, which is used for
Quality inspection is advised in conjunction when through financial products such as internet sales insurance, financings.The application environment includes server-side and client,
Wherein, it is attached between server-side and client by network, which can be cable network or wireless network, client
End is specifically including but not limited to various personal computers, laptop, smart phone and tablet computer and portable wearable
Equipment, server-side can specifically be realized with the server cluster that independent server or multiple servers form.Server-side is
The user of financial product seller, client is financial product purchaser, when the application that user passes through client
When (Application, APP) carries out financial product on-line purchase, server-side starts self-service double record processes, and records at self-service pair
The relevant information that user is obtained in journey is completed to close rule quality inspection in real time.
In one embodiment, as shown in Fig. 2, providing a kind of double record quality detecting methods, the service in Fig. 1 is applied in this way
It is illustrated for end, details are as follows:
S1: if receiving the financial product purchase request of client transmission, sending the double record instructions of starting to client,
In, the double record instructions of the starting make client open cam device and microphone apparatus.
Specifically, user can apply for the financial products such as purchase insurance, financing by the APP of client, when user is in APP
Purchase interface submit purchase application after, client to server-side send financial product purchase request, the financial product purchase ask
The identification information of the registration information comprising user and financial product to be purchased is sought, registration information, which can specifically include, to succeed in registration
Username and password etc..
Server-side receives the financial product purchase request of client transmission, obtains user from financial product purchase request
Registration information, and to the registration information carry out validity check, if being inquired in legitimate user's information database of prediction
The registration information then confirms the validity verification and passes through, if being to inquire the registration in preset legitimate user's information database
Information then confirms the validity authentication failed.
If validation verification passes through, server-side confirms that user is legal, allow user carry out financial product purchase into one
Step operation, server-side send the double record instructions of starting to client, and the double record instructions of the starting are used to indicate client and open camera
Device and microphone apparatus, to start the process of double records.Wherein, cam device is used to record facial video or the bat of user
The face-image of user is taken the photograph, microphone apparatus is used to record the voice of user.
If validation verification fails, server-side refuses the purchase request of this financial product, and registers to client feedback
The invalid prompt information of information.
S2: if receiving double record start completion message of client return, according to preset selection mode, from preset
It is chosen in problem base and closes rule problem, and send client for voice the problem of closing rule problem by way of machine talk.
Specifically, client receives the double record instructions of starting of server-side transmission, carries out the preparation of double records, completes out
Cam device and microphone apparatus are opened, and the display interface in APP is shown after recording frame and double record prompt informations, to service
End returns to double record start completion message.
Server-side receives double record start completion message of client transmission, confirms that double record preparations are completed, Ke Yikai
Begin to carry out double records, then chooses rule problem of closing from preset problem base according to preset selection mode, and pass through machine talk
Mode is converted to problem voice for rule problem is closed, and sends client for the problem voice.
Wherein, preset selection mode specifically can be according to the finance to be purchased for including in financial product purchase request
The identification information of product chooses preset conjunction rule problem relevant to the financial product to be purchased from preset problem base,
When preset conjunction rule more problems relevant to the financial product to be purchased, the conjunction of preset quantity can be therefrom randomly choosed
Rule problem;Preset selection mode can also be the conjunction rule problem that preset quantity is randomly choosed directly from preset problem base.
It should be noted that preset selection mode can be configured according to the needs of practical application, do not limit herein
System.
It is previously stored with the various different insurance relevant conjunction rule problems of type in preset problem base and each conjunction rule are asked
Inscribe corresponding conjunction rule answer.It closes rule problem and refers in financial product sales process to be to avoid sale from misleading and ensure to buy to use
Family is true legal and the problem of be arranged, be included in sales process and whether informed user's financial product title, undertaking company
Title, pay charge way after the hesitation phase the problems such as surrender loss risk, and require user to provide the problems such as legal identity proves.
The each conjunction selected is advised problem by server-side, is successively converted to problem voice, and ask one according to one for problem voice
The problem of mode answered is sent in sequence to client, i.e., sends a conjunction rule problem every time voice, when the user of client answers
After, the problem of retransmiting next conjunction rule problem voice, until answer until all closing rule problems.
After client receives the problem of conjunction rule problem of server-side transmission voice, this is played by voice playing device and is asked
Voice is inscribed, and answers the answer voice of conjunction rule problem using microphone apparatus acquisition user, meanwhile, it is clapped by cam device
Take the photograph video image of the user when answering conjunction rule problem.
Collected answer voice and the video image of shooting correspondence are sent server-side by client.
S3: the answer voice and video image that client is sent is received, wherein vocal answer is that client passes through microphone
The user of device acquisition answers the voice for closing rule problem, and video image is that client is answered by the user that cam device is shot
Close the video of rule problem.
Specifically, the user that server-side receives that client is sent answers answer voice and video image when closing rule problem,
And rule problem, answer voice and video image will be closed and carry out corresponding preservation.
S4: video image is inputted into preset human face recognition model and carries out recognition of face, obtains face recognition result.
Specifically, preset human face recognition model is used to carry out face match cognization to video image, to confirm user's
Whether identity is legal.The human face recognition model extracts the facial image of user from the video image of input, and by the face figure
Similarity mode is carried out between the facial image in identity information that picture is provided previously with user, and according to the knot of similarity mode
Fruit determines face recognition result, and face recognition result is that the identity of user is legal if similarity mode success, if similarity
With failure, then face recognition result is that the identity of user is illegal.
Wherein, the identity information that user is provided previously specifically can be ID card information, and server-side is uploaded according to user
Identity card picture or video containing identity card, are extracted and preserved the facial image in identity card.
Wherein, in human face recognition model, to the facial image in the facial image and identity information in video image it
Between carry out similarity mode process can specifically include:
According to identical feature extraction mode, the image feature vector of the facial image in video image is extracted, as the
One feature vector, and the image feature vector of the facial image in identity information is extracted, as second feature vector;
According to preset calculation, the characteristic similarity between first eigenvector and second feature vector is calculated;
If characteristic similarity is greater than or equal to preset similarity threshold, the facial image and body in video image are confirmed
Similarity mode success between facial image in part information;
If characteristic similarity is less than preset similarity threshold, the facial image and identity information in video image are confirmed
In facial image between similarity mode failure.
S5: if face recognition result is legal for the identity of user, preset speech recognition modeling, answer case voice are used
Text conversion is carried out, the corresponding answer text of answer voice is obtained.
Specifically, if it is legal according to the identity that the face recognition result that step S4 is obtained is user, further to user
Answer the real-time quality inspection of answer progress for closing rule problem.
Preset speech recognition modeling inputs language for converting speech into text, by answer voice obtained in step S2
Sound identification model, speech recognition modeling answer case voice carry out speech recognition, the corresponding answer text of output answer voice.
Wherein, preset speech recognition modeling specifically can be using based on hidden Markov model (Hidden Markov
Model, HMM) speech recognition algorithm, can also using based on by gauss hybrid models (Gaussian Mixed Model,
GMM) and the speech recognition algorithm of GMM-HMM model that is composed of hidden Markov model, but it is not limited to this, in the present invention
In embodiment with no restrictions to the specific implementation algorithm of speech recognition modeling.
S6: it is obtained from preset problem base and closes the corresponding conjunction rule answer of rule problem, and answer text and conjunction rule are answered
Case is matched, and matching result is obtained.
Specifically, it is prestored in preset problem base and is stored with the corresponding conjunction rule answer of each conjunction rule problem, that is, met and close
Advise the answer that quality inspection requires.
The answer text that step S5 is obtained is matched with conjunction rule answer, matched mode specifically can be to answer
Text is compared with closing rule answer and carrying out text character, if answer text reaches pre- with the quantity for closing same text character in rule answer
If amount threshold, then confirm matching result be successful match, otherwise, if answer text and close rule answer in same text character
Quantity be not up to preset amount threshold, then confirm matching result be it fails to match.
Matched mode can also be that the calculation method based on statistical calculation method or based on semantic understanding calculates
Text similarity between answer text and conjunction rule answer, if text similarity is greater than or equal to preset similarity threshold,
Confirmation confirmation matching result is successful match, otherwise, if text similarity is less than preset similarity threshold, confirms matching knot
Fruit is that it fails to match.
Wherein, the text similarity being calculated based on statistical calculation method can be answer text vector and close rule
Hamming distance between cosine similarity or answer text between text vector and conjunction rule answer, based on semantic understanding
Calculation method according between concept in Semantic hierarchy relationship in preset dictionary hyponymy or synonymy carry out
It calculates.
S7: if matching result is successful match, confirm that conjunction rule quality inspection passes through, and the finger for terminating double records is sent to client
It enables.
Specifically, if the matching result that step S6 is obtained is successful match, rule problem and user are closed in server-side confirmation
Answer meet close rule quality inspection requirement, i.e., confirmation close rule quality inspection pass through;If matching result is that it fails to match, confirm that conjunction rule are asked
The answer of topic or user are unsatisfactory for closing the requirement of rule quality inspection, i.e. confirmation is closed rule quality inspection and do not passed through.
Pass through if rule quality inspection is closed in server-side confirmation, the instruction for terminating double records is sent to client, client receives this
Cam device and microphone apparatus are closed after instruction, terminates double record processes, and are continued according to the instruction of server-side subsequent
Financial product process of purchase.
It should be noted that the quantity for the conjunction rule problem that need to be answered due to double record process users may be one is also likely to be
It is multiple, therefore, when the quantity for closing rule problem is greater than one, it can limit only when the corresponding matching result of each conjunction rule problem
It is that successfully just the quality inspection of confirmation conjunction rule passes through, can also limits when the corresponding matching result of conjunction rule problem accounts for conjunction as successful quantity
Confirmation is closed when the ratio of the quantity of rule problem reaches preset proportion threshold value passes through between rule.
Do not pass through if rule quality inspection is closed in server-side confirmation, terminates the process of purchase of this financial product, and return to client
Bout rule quality inspection does not pass through the prompt information that can not continue to buy financial product.
In the present embodiment, in the sales process of internet financial product, if server-side receives the gold of client transmission
Melt product purchase request, then sends the double record instructions of starting to start double records, if receiving double records of client return to client
Start completion message is chosen from preset problem base then according to preset selection mode and closes rule problem, pass through machine talk
Voice the problem of closing rule problem is sent client by mode, and receives the answer language of user's pairing rule problem of client return
Sound and user answer video image when closing rule problem, and server-side carries out face knowledge to video image by human face recognition model
Not, after confirming that the identity of user is legal, the corresponding answer text of answer voice of user is identified by speech recognition modeling, and
The conjunction rule answer corresponding with rule problem is closed of answer text is matched, confirms that conjunction rule quality inspection passes through if successful match.It realizes
Double record processes are automatically performed by way of self-service double records, and the content of double records carried out to close in real time automatically and advises quality inspection, from
And double records and conjunction rule quality inspection is made not to need manpower intervention, human cost is effectively reduced, and improve the efficiency for closing rule quality inspection.
In one embodiment, as shown in figure 3, in step s 6, it is corresponding that conjunction rule problem is obtained from preset problem base
Rule answer is closed, and answer text is matched with rule answer is closed, matching result is obtained, specifically comprises the following steps:
S61: answer case text carries out keyword extraction, obtains the answer keyword that answer text includes.
Specifically, using preset word segmentation processing algorithm, answer case text carries out word segmentation processing, and obtaining answer text includes
Several words.Wherein, word segmentation processing algorithm can specifically use the segmenting method based on string matching, or use and be based on
The full cutting method of statistical language model, but it is not limited to this, does not do have to the mode of word segmentation processing in embodiments of the present invention
Body limitation.
Wherein, character string is carried out according to certain scanning strategy answer case text based on the segmenting method of string matching to cut
Point, and the character string that cutting obtains is matched one by one with the entry in preset dictionary, if finding the entry in dictionary,
Then successful match.According to the difference of scanning strategy, the segmenting method based on string matching can be divided into positive matching, reverse
Match and the different modes such as bi-directional matching.Full cutting method based on statistical language model is syncopated as and preset dictionary first
In all possible word that matches of entry, then determine optimal cutting as a result, its advantages exist with statistical language model
In can solve participle in ambiguity problem.
After obtaining the corresponding several words of answer text, use the machine learning model of supervision for each Word prediction
The weighted score of one [0,1], the more big then word importance of the weighted score of word are higher.There is the machine learning model of supervision
Can using training data from extract method carry out model training, i.e., from automatic mining training data in history answer text into
Row model training carries out feature extraction to training data, and logic-based regression algorithm analyzes the feature extracted, in advance
The importance of each text string in training data is surveyed, meanwhile, there is the machine learning model of supervision can also be according to the answer of user
Text progress self study is perfect constantly to carry out model, improves the accuracy of prediction.
According to the weighted score of each word, the word that weighted score is greater than preset score threshold is determined as answer and is closed
Key word.
S62: it is obtained from preset problem base and closes the corresponding conjunction rule answer of rule problem, and answer keyword and conjunction rule are answered
Case carry out text character matching, if close rule answer in be matched to each answer keyword, confirm matching result for matching at
Function.
Specifically, it is obtained from preset problem base and closes the corresponding conjunction rule answer of rule problem, closed rule answer and be set in advance
And it is preset as in text library with corresponding be stored in of rule problem is closed, and close rule answer and can according to need progress dynamic adjustment.
The each answer keyword and conjunction rule answer that step S61 is obtained carry out text character matching, are closed and are advised by traversal
Each text character in answer inquires the text character that whether there is answer keyword in closing rule answer, if answering closing rule
There are the text characters of answer keyword in case, then confirmation is matched to the answer keyword in closing rule answer.
It is queried to when the text character of each answer keyword can advise in answer in conjunction, then confirm that matching result is
With success.
If the text character for some answer keyword that answer text includes is not queried in closing rule answer, continue
Step S63 is executed, is further matched.
S63: closing the answer keyword that is not matched to of rule answer if it exists, then using the answer keyword not being matched to as to
Identidication key, and obtain from preset near synonym dictionary the near synonym of each keyword to be identified.
Specifically, if the text character for some answer keyword that answer text includes is not queried in closing rule answer
It arrives, then using the answer keyword not being queried to as keyword to be identified, and obtaining from preset near synonym dictionary should
The corresponding near synonym of keyword to be identified.
Wherein, pre-saved in preset near synonym dictionary in problem base conjunction rule problem and close advise answer it is relevant
Various key words and its corresponding near synonym set.
Keyword to be identified is inquired in the key words for including near synonym dictionary, if inquiring, obtaining should be wait know
Near synonym in the corresponding near synonym set of other keyword.
It should be understood that the corresponding near synonym of keyword to be identified got are either one or more.
S64: by the near synonym of each keyword to be identified and rule answer progress text character matching is closed, if each to be identified
At least one near synonym and conjunction rule answer matches of keyword, then confirm that matching result is successful match.
Specifically, the near synonym of the obtained keyword to be identified of step S63 and conjunction rule answer are subjected to text character matching,
Its matching way can match the identical method of answer keyword in closing rule answer using with step S63.
For each keyword to be identified, if at least one near synonym of the keyword to be identified and conjunction rule answer matches,
Then confirm the keyword to be identified and close rule answer matches success, if each near synonym of the keyword to be identified are answered with conjunction rule
Case mismatches, then confirms that it fails to match in the keyword to be identified and conjunction rule answer.
If each keyword to be identified with conjunction rule answer matches success, confirms that matching result is successful match, if depositing
In keyword to be identified and rule answer matches failure is closed, then confirms that matching result is that it fails to match.
For example, if close rule problem be " whether user it is known that detailed contract terms? ", corresponding conjunction rule answer is
" knowing ", and the near synonym set comprising words such as " knowing ", " knowing " and " clear " is provided near synonym dictionary.That
, when user answers " knowing ", " knowing " or " clear ", obtaining matching result is successful match.
S65: if each near synonym of keyword to be identified and conjunction rule answer mismatch, confirm matching result for matching mistake
It loses.
Specifically, if obtaining according to step S64 there are keyword to be identified and closing rule answer matches failure, confirm matching
As a result for it fails to match.
In the present embodiment, during answer text and conjunction rule answer are carried out matched, answer case text is carried out first
Keyword extraction obtains answer keyword, answer keyword and conjunction rule answer is then carried out text character matching, if advising closing
It is matched to each answer keyword in answer and then confirms successful match, if certain answer keywords are not matched in conjunction rule answer
It arrives, then further obtains the near synonym of each answer keyword by preset near synonym dictionary, using the near synonym and close rule
Answer carries out text character matching, if being matched at least one near synonym of each answer keyword in closing rule answer, really
Recognize successful match, this to be matched in such a way that text matches and near synonym analysis combine, computation complexity is low, effectively
Execution efficiency is improved, while improving the accuracy of matching result.
In one embodiment, as shown in figure 4, in the step s 7, if matching result is successful match, confirming conjunction rule quality inspection
Pass through, and send the instruction for terminating double records to client, specifically comprises the following steps:
S71: if matching result is successful match, face is carried out to video image using preset micro- Expression Recognition model
Emotion identification obtains user and is answering emotional state when closing rule problem.
Specifically, micro- Expression Recognition model is used to carry out facial emotions identification to the facial image in video image, obtains
Facial image corresponds to the probability value of preset a variety of moods, and the maximum mood of probability value is confirmed as to the mood of facial image
State, i.e. user are answering emotional state when closing rule problem.
For example, preset a variety of moods in micro- table Expression Recognition model can be set as glad, sad, surprised, light
Slight, fear, indignation and detesting etc., can acquire in advance and respectively represent the great amount of samples pictures of these moods and be labeled, be formed
Then samples pictures collection is trained using convolutional neural networks model or classifier, finally obtain micro- Expression Recognition model.
S72: it is obtained from preset problem base and closes the corresponding preset reasonable emotional state of rule problem.
Specifically, the corresponding reasonable emotional state of each conjunction rule problem has been pre-saved in preset problem base, rationally
Emotional state is either one or more.
Difference closes the corresponding reasonable emotional state of rule problem can be identical or not identical, and each conjunction rule problem is corresponding
Reasonable emotional state is related to the conjunction rule content of problem.
S73: if emotional state of the user when answering conjunction rule problem is consistent with reasonable emotional state, confirm conjunction rule quality inspection
Pass through, and sends the instruction for terminating double records to client.
Specifically, if emotional state of the user obtained according to step S71 when answering conjunction rule problem is obtained with step S72
The corresponding reasonable emotional state of conjunction rule problem it is identical, alternatively, user belongs to this answering emotional state when closing rule problem
Close one of corresponding multiple reasonable emotional states of rule problem, then confirm user answer emotional state when closing rule problem with
Reasonable emotional state is consistent, i.e., user emotional state it is normal.
According to step S6 obtain matching result be successful match on the basis of, if the emotional state of user is normal, recognize
Surely duplicity quality inspection is not present, i.e. confirmation is closed rule quality inspection and passed through, and the instruction for terminating double records is sent to client, client receives
Cam device and microphone apparatus are closed after the instruction, terminates double record processes, and after continuing according to the instruction of server-side
The process of purchase of continuous financial product.
It should be noted that the quantity for the conjunction rule problem that need to be answered due to double record process users may be one is also likely to be
It is multiple, therefore, when close rule problem quantity be greater than one when, can limit only when user answer it is each close rule problem when
Emotional state reasonable emotional state corresponding with conjunction rule problem is consistent, i.e. mood of the user when answering each conjunction rule problem
State is normal, and just confirmation is closed rule quality inspection and passed through.
In the present embodiment, on the basis of matching result is successful match, micro- Expression Recognition model is further used to view
Frequency image carries out facial emotions identification, obtains emotional state of the user when answering the conjunction rule problem, and pass through the mood
State preset reasonable emotional state corresponding with rule problem is closed is compared, and judges that user is answering mood when closing rule problem
Whether state is normal, and when emotional state of the user when answering conjunction rule problem is normal, just confirmation is closed rule quality inspection and passed through, and realizes
It carries out closing rule quality inspection in conjunction with micro- Expression Recognition, to avoid the risk of duplicity quality inspection, improves the accuracy for closing rule quality inspection result.
In one embodiment, as shown in figure 5, in step S71, using preset micro- Expression Recognition model to video image
Facial emotions identification is carried out, user is obtained and is answering emotional state when closing rule problem, specifically comprise the following steps:
S711: video frame extraction is carried out to video image according to preset extracting mode, obtains target frame image.
Specifically, preset extracting mode, which can be, extracts a frame at interval of preset frame number, is also possible to extract at random
Several frames can be specifically configured, herein with no restrictions according to the needs of practical application.
S712: target frame image is inputted into preset micro- Expression Recognition model, obtains user in each preset micro- expression
Probability under state.
Specifically, preset micro- Expression Recognition model carries out human facial feature extraction to each target frame image of input, and
Micro- emotional state that user is identified according to expressive features, obtains each target frame image under N number of preset micro- emotional state
Elementary probability, wherein N is positive integer.
Wherein, preset micro- emotional state includes a variety of slight expression states such as happiness, anger, grief and joy, such as: feel puzzled, anger
Up to 54 kinds of burning, greatly surprised, dog-tired etc. is mutually with micro- expressions of nuance in fire.
Due to target frame image may have it is multiple, obtaining each target frame image under N number of micro- emotional state
After elementary probability, with micro- emotional state be calculate dimension, calculate under each micro- emotional state each target frame image it is basic
The average value or weighted average of probability, obtain the combined chance of each micro- emotional state, which is that user exists
Probability under micro- emotional state.
For example, if the quantity of target frame image has 3 frames, for preset micro- emotional state A.First frame is in micro- emotional state
Elementary probability under A is 30%, and elementary probability of second frame at micro- emotional state A is 28%, and third frame is in micro- emotional state A
Under elementary probability be 25%, then calculate at micro- emotional state A the elementary probability of 3 frame images average value, i.e. (30%+
20%+25%)/3=27.7%, the combined chance for obtaining micro- emotional state A is 27.7%, i.e., user is at micro- emotional state A
Probability be 27.7%.
S713: the probability for being greater than preset probability threshold value is chosen from each probability, and the probability selected is corresponding micro-
Emotional state is as the micro- emotional state of target.
Specifically, from N number of probability that step S712 is obtained, the probability that probability value is greater than preset probability threshold value is corresponding
Micro- emotional state as the micro- emotional state of target.
S714: according to corresponding relationship preset between micro- emotional state and emotional state, by the way of ballot, statistics is every
The poll of the corresponding emotional state of a micro- emotional state of target, and rule are closed answering using the most emotional state of poll as user
Emotional state when problem.
In the present embodiment, due to only embodying the slight change of expression between different micro- emotional states, in micro- expression
When the quantity of state is more, the corresponding relationship between micro- emotional state and emotional state can be pre-established, by numerous micro- expressions
State classification is concluded into the emotional state of macroscopic view, and has apparent mood difference between different emotional states.For example, micro-
Emotional state may include up to 54 kinds expression happiness, anger, grief and joy and the phase such as feel puzzled, make one's blood boil, is greatly surprised, is dog-tired
Mutually with micro- expression of nuance, emotional state may include it is glad, sad, nervous, scorn, fear, the moods such as indignation,
In, micro- emotional state such as " beaming with smiles ", " in high spirits " can correspond to " happiness " this emotional state.
Specifically, according to the corresponding relationship between micro- emotional state and emotional state, the micro- emotional state of each target is obtained
Corresponding emotional state, and by way of ballot, it votes for emotional state, counts the poll that each emotional state obtains.Example
Such as, the micro- emotional state of target includes A1, A2 and A3, and emotional state includes K1 and K2, if the corresponding emotional state of A1 is K1, A2 pairs
The emotional state answered is K1, and the corresponding emotional state of A3 is K2, then the poll of emotional state K1 is 2 tickets, the ticket of emotional state K2
Number is 1 ticket.
From the poll of each emotional state, the most emotional state of selection poll is as user when answering conjunction rule problem
Emotional state.
In the present embodiment, facial emotions identification is being carried out to video image using micro- Expression Recognition model, user is being obtained and exists
When answering emotional state when closing rule problem, by carrying out micro- Expression Recognition to video image, user is obtained each preset
After probability under micro- emotional state, the corresponding micro- emotional state of probability for being greater than preset probability threshold value is chosen as the micro- table of target
Situation state, and by the way of ballot, the poll of the corresponding emotional state of the micro- emotional state of each target is counted, poll is most
Emotional state as user answer close rule problem when emotional state, to realize through user in each micro- expression shape
Emotional state of the probability accurate judgement user when answering conjunction rule problem under state, the recognition result of emotional state accurately may be used
It leans on, whether to close rule quality inspection by providing reliable judgment basis.
In one embodiment, further include the unsanctioned judgement processing of pairing rule quality inspection, details are as follows after step S72:
If emotional state and reasonable emotional state of the user when answering conjunction rule problem are inconsistent, the quality inspection of conjunction rule is confirmed not
Pass through, and returns to abnormal prompt information to client.
Specifically, if emotional state of the user obtained according to step S71 when answering conjunction rule problem is obtained with step S72
The corresponding reasonable emotional state of conjunction rule problem it is not identical, alternatively, user does not belong to answering emotional state when closing rule problem
In any one of corresponding multiple reasonable emotional states of conjunction rule problem, then confirm that user is answering feelings when closing rule problem
Not-ready status and reasonable emotional state are inconsistent, i.e., user emotional state it is abnormal.
On the basis of according to step S6, to obtain matching result be successful match, if the emotional state of user is abnormal, though
Rule problem and the requirement of user answered satisfaction and close rule quality inspection are closed in right server-side confirmation, but since user is answering conjunction rule problem
When emotional state exist abnormal, therefore, still assert that there may be the risk of duplicity quality inspection, i.e. confirmation is closed rule quality inspection and do not led to
Cross, and terminate the process of purchase of this financial product, at the same to client return close rule quality inspection do not pass through can not continue purchase gold
Melt the prompt information of product.
In the present embodiment, if emotional state and reasonable emotional state of the user when answering conjunction rule problem are inconsistent, immediately
Rule problem is closed in server-side confirmation and the answer of user meets the requirement for closing rule quality inspection, still assert that there may be duplicity quality inspections
Risk, i.e., confirmation close rule quality inspection do not pass through, the process of purchase of this financial product is terminated, to avoid the wind of duplicity quality inspection
The accuracy for closing rule quality inspection result is improved in danger.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process
Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit
It is fixed.
In one embodiment, a kind of double record quality inspection devices are provided, this pair records double record matter in quality inspection device and above-described embodiment
Detecting method corresponds.As shown in fig. 6, this pair record quality inspection device include: starting module 61, choose module 62, receiving module 63,
Module 67 is completed in identification module 64, conversion module 65, matching module 66 and quality inspection.Detailed description are as follows for each functional module:
Starting module 61, if the financial product purchase request for receiving client transmission, is opened to client transmission
Dynamic double record instructions, wherein the double record instructions of the starting make client open cam device and microphone apparatus;
Module 62 is chosen, if double record start completion message for receiving client return, according to preset selection
Mode is chosen from preset problem base and closes rule problem, and is sent out voice the problem of closing rule problem by way of machine talk
It is sent to client;
Receiving module 63, for receiving the answer voice and video image of client transmission, wherein vocal answer is client
The voice for closing rule problem is answered at end by the user that microphone apparatus acquires, and video image is that client is clapped by cam device
The user taken the photograph answers the video for closing rule problem;
Identification module 64 carries out recognition of face for video image to be inputted preset human face recognition model, obtains face
Recognition result;
Conversion module 65 uses preset speech recognition mould if the identity for face recognition result to be user is legal
Type, answer case voice carry out text conversion, obtain the corresponding answer text of answer voice;
Matching module 66 closes the corresponding conjunction rule answer of rule problem for obtaining from preset problem base, and answer is literary
This is matched with conjunction rule answer, obtains matching result;
Module 67 is completed in quality inspection, if being successful match for matching result, confirms that conjunction rule quality inspection passes through, and to client
Send the instruction for terminating double records.
Further, matching module 66 includes:
Extracting sub-module 661 carries out keyword extraction for answering case text, it is crucial to obtain the answer that answer text includes
Word;
Successful match submodule 662 closes the corresponding conjunction rule answer of rule problem for obtaining from preset problem base, and will
Answer keyword and conjunction rule answer carry out text character matching, if being matched to each answer keyword in closing rule answer, really
Recognizing matching result is successful match;
Near synonym acquisition submodule 663, the answer keyword not being matched to for closing rule answer if it exists, then will not match
The answer keyword arrived obtains the close of each keyword to be identified as keyword to be identified from preset near synonym dictionary
Adopted word;
Near synonym matched sub-block 664, for the near synonym of each keyword to be identified and conjunction rule answer to be carried out text
A character match, if at least one near synonym of each keyword to be identified and conjunction rule answer matches, confirm that matching result is
With success;
It fails to match submodule 665, if for keyword to be identified each near synonym and close rule answer and mismatch, really
Recognizing matching result is that it fails to match.
Further, quality inspection completion module 67 includes:
Emotion identification submodule 671 uses preset micro- Expression Recognition model if being successful match for matching result
Facial emotions identification is carried out to video image, user is obtained and is answering emotional state when closing rule problem;
Reasonable mood acquisition submodule 672 closes the corresponding preset conjunction of rule problem for obtaining from preset problem base
Manage emotional state;
Successful match submodule 673, if emotional state and reasonable mood for user when answering the conjunction rule problem
State consistency then confirms that conjunction rule quality inspection passes through, and sends the instruction for terminating double records to client.
Further, Emotion identification submodule 671 includes:
Extraction unit 6711 obtains target for carrying out video frame extraction to video image according to preset extracting mode
Frame image;
Recognition unit 6712 obtains user each for target frame image to be inputted preset micro- Expression Recognition model
Probability under preset micro- emotional state;
Screening unit 6713 for choosing the probability for being greater than preset probability threshold value from each probability, and will be selected
The corresponding micro- emotional state of probability is as the micro- emotional state of target;
Statistic unit 6714 is used for according to corresponding relationship preset between micro- emotional state and emotional state, using ballot
Mode, count the poll of the corresponding emotional state of the micro- emotional state of each target, and using the most emotional state of poll as
User is answering emotional state when closing rule problem.
Further, module 67 is completed in quality inspection further include:
It fails to match submodule 674, if answering emotional state and reasonable emotional state when closing rule problem for user
It is inconsistent, then confirm that conjunction rule quality inspection does not pass through, and return to abnormal prompt information to client.
Specific about double record quality inspection devices limits the restriction that may refer to above for double record quality detecting methods, herein not
It repeats again.Modules in above-mentioned double record quality inspection devices can be realized fully or partially through software, hardware and combinations thereof.On
Stating each module can be embedded in the form of hardware or independently of in the processor in computer equipment, can also store in a software form
In memory in computer equipment, the corresponding operation of the above modules is executed in order to which processor calls.
In one embodiment, a kind of computer equipment is provided, which can be server, internal structure
Figure can be as shown in Figure 7.The computer equipment includes processor, the memory, network interface sum number connected by system bus
According to library.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory of the computer equipment includes
Non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data
Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The network interface of machine equipment is used to communicate with external terminal by network connection.When the computer program is executed by processor with
Realize a kind of double record quality detecting methods.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory simultaneously
The computer program that can be run on a processor, processor realize double record quality inspection sides in above-described embodiment when executing computer program
The step of method, such as step S1 shown in Fig. 2 to step S7.Alternatively, processor realizes above-mentioned implementation when executing computer program
The function of each module/units of double record quality inspection devices in example, such as module 61 shown in Fig. 6 is to the function of module 67.To avoid weight
Multiple, details are not described herein again.
In one embodiment, a kind of computer readable storage medium is provided, computer program, computer are stored thereon with
Double record quality detecting methods in above method embodiment are realized when program is executed by processor, alternatively, the computer program is by processor
The function of each module/unit in double record quality inspection devices in above-mentioned apparatus embodiment is realized when execution.To avoid repeating, herein no longer
It repeats.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram
(DDRSDRAM), enhanced SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory
Bus (Rambus) directly RAM (RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic
RAM (RDRAM) etc..
It is apparent to those skilled in the art that for convenience of description and succinctly, only with above-mentioned each function
Can unit, module division progress for example, in practical application, can according to need and by above-mentioned function distribution by different
Functional unit, module are completed, i.e., the internal structure of described device is divided into different functional unit or module, more than completing
The all or part of function of description.
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations;Although referring to aforementioned reality
Applying example, invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each
Technical solution documented by embodiment is modified or equivalent replacement of some of the technical features;And these are modified
Or replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all
It is included within protection scope of the present invention.
Claims (10)
1. a kind of double record quality detecting methods, which is characterized in that recording quality detecting methods for described pair includes:
If receiving the financial product purchase request of client transmission, the double record instructions of starting are sent to the client, wherein
The double record instructions of starting make the client open cam device and microphone apparatus;
If receiving double record start completion message that the client returns to ask according to preset selection mode from preset
It is chosen in exam pool and closes rule problem, and send the client for voice the problem of the conjunction rule problem by way of machine talk
End;
Receive the answer voice and video image that the client is sent, wherein the vocal answer is that the client passes through
The user of the microphone apparatus acquisition answers the voice for closing rule problem, and the video image is that the client passes through institute
The user for stating cam device shooting answers the video for closing rule problem;
The video image is inputted into preset human face recognition model and carries out recognition of face, obtains face recognition result;
If the face recognition result is that the identity of the user is legal, preset speech recognition modeling is used, is answered described
Case voice carries out text conversion, obtains the corresponding answer text of the answer voice;
Obtain the conjunction rule problem corresponding conjunction rule answer from the preset problem base, and by the answer text with it is described
It closes rule answer to be matched, obtains matching result;
If the matching result is successful match, confirm that conjunction rule quality inspection passes through, and send the double records of end to the client
Instruction.
2. double record quality detecting methods as described in claim 1, which is characterized in that described to obtain institute from the preset problem base
It states and closes the corresponding conjunction rule answer of rule problem, and the answer text is matched with conjunction rule answer, obtain matching result
Include:
Keyword extraction is carried out to the answer text, obtains the answer keyword that the answer text includes;
Obtain the conjunction rule problem corresponding conjunction rule answer from the preset problem base, and by the answer keyword and institute
It states conjunction rule answer and carries out text character matching, if being matched to each answer keyword in conjunction rule answer, confirm
The matching result is successful match;
The answer keyword for closing rule answer and not being matched to if it exists, then make the answer keyword not being matched to
For keyword to be identified, and obtain from preset near synonym dictionary the near synonym of each keyword to be identified;
The near synonym of each keyword to be identified and conjunction rule answer are subjected to text character matching, if each institute
At least one the described near synonym and the conjunction for stating keyword to be identified advise answer matches, then confirm the matching result for matching
Success;
If each of the keyword to be identified near synonym and conjunction rule answer mismatch, the matching result is confirmed
For it fails to match.
3. double record quality detecting methods as described in claim 1, which is characterized in that if the matching result is successful match,
Then confirm that conjunction rule quality inspection passes through, and includes: to the instruction that the client sends the double records of end
If the matching result is successful match, face is carried out to the video image using preset micro- Expression Recognition model
Emotion identification obtains emotional state of the user when answering the conjunction rule problem;
The corresponding preset reasonable emotional state of the conjunction rule problem is obtained from the preset problem base;
If emotional state of the user when answering the conjunction rule problem is consistent with the reasonable emotional state, conjunction rule are confirmed
Quality inspection passes through, and the instruction for terminating double records is sent to the client.
4. double record quality detecting methods as claimed in claim 3, which is characterized in that described to use preset micro- Expression Recognition model pair
The video image carries out facial emotions identification, obtains emotional state of the user when answering the conjunction rule problem and includes:
Video frame extraction is carried out to the video image according to preset extracting mode, obtains target frame image;
The target frame image is inputted into preset micro- Expression Recognition model, obtains the user in each preset micro- table
Probability under situation state;
The probability for being greater than preset probability threshold value is chosen from each probability, and the probability selected is corresponding described
Micro- emotional state is as the micro- emotional state of target;
According to preset corresponding relationship between micro- emotional state and the emotional state, by the way of ballot, statistics is every
The poll of the corresponding emotional state of a micro- emotional state of the target, and the most emotional state of the poll is made
For emotional state of the user when answering the conjunction rule problem.
5. double record quality detecting methods as described in claim 3 or 4, which is characterized in that described from the preset problem base
After obtaining the corresponding preset reasonable emotional state of the conjunction rule problem, double record quality detecting methods further include:
If emotional state and the reasonable emotional state of the user when answering the conjunction rule problem are inconsistent, conjunction is confirmed
Rule quality inspection does not pass through, and returns to abnormal prompt information to the client.
6. a kind of double record quality inspection devices, which is characterized in that recording quality inspection devices for described pair includes:
Starting module, if the financial product purchase request for receiving client transmission, sends to the client and start
Double record instructions, wherein the double record instructions of starting make the client open cam device and microphone apparatus;
Module is chosen, if the double record start completion message returned for receiving the client, according to preset selection side
Formula is chosen from preset problem base and closes rule problem, and by voice the problem of the conjunction rule problem by way of machine talk
It is sent to the client;
Receiving module, the answer voice and video image sent for receiving the client, wherein the vocal answer is institute
It states client and answers the voice for closing rule problem by the user that the microphone apparatus acquires, the video image is described
Client answers the video for closing rule problem by the user that the cam device is shot;
Identification module carries out recognition of face for the video image to be inputted preset human face recognition model, obtains face knowledge
Other result;
Conversion module uses preset speech recognition if the identity for the face recognition result to be the user is legal
Model carries out text conversion to the answer voice, obtains the corresponding answer text of the answer voice;
Matching module, for obtaining the corresponding conjunction rule answer of the conjunction rule problem from the preset problem base, and will be described
Answer text is matched with conjunction rule answer, obtains matching result;
Module is completed in quality inspection, if being successful match for the matching result, confirms that conjunction rule quality inspection passes through, and to the client
End sends the instruction for terminating double records.
7. record quality inspection devices as claimed in claim 6 double, which is characterized in that the matching module includes:
Extracting sub-module obtains the answer that the answer text includes and closes for carrying out keyword extraction to the answer text
Key word;
Successful match submodule, for obtaining the corresponding conjunction rule answer of the conjunction rule problem from the preset problem base, and
The answer keyword and conjunction rule answer are subjected to text character matching, if being matched to each institute in conjunction rule answer
Answer keyword is stated, then confirms that the matching result is successful match;
Near synonym acquisition submodule closes the answer keyword that is not matched to of rule answer for described if it exists, then will not
The answer keyword being fitted on is obtained from preset near synonym dictionary each described to be identified as keyword to be identified
The near synonym of keyword;
Near synonym matched sub-block, for carrying out the near synonym of each keyword to be identified and conjunction rule answer
Text character matching, if answer matches are advised at least one described near synonym of each keyword to be identified and the conjunction,
Confirm that the matching result is successful match;
It fails to match submodule, if not advising answer not for each of the keyword to be identified near synonym and conjunctions
Match, then confirms that the matching result is that it fails to match.
8. double record quality inspection devices as claimed in claim 6, which is characterized in that the quality inspection completes module and includes:
Emotion identification submodule uses preset micro- Expression Recognition model pair if being successful match for the matching result
The video image carries out facial emotions identification, obtains emotional state of the user when answering the conjunction rule problem;
Reasonable mood acquisition submodule, for obtaining the corresponding preset conjunction of the conjunction rule problem from the preset problem base
Manage emotional state;
Mood matched sub-block, if emotional state and the reasonable mood for the user when answering the conjunction rule problem
State consistency then confirms that conjunction rule quality inspection passes through, and sends the instruction for terminating double records to the client.
9. a kind of computer equipment, including memory, processor and storage are in the memory and can be in the processor
The computer program of upper operation, which is characterized in that the processor realized when executing the computer program as claim 1 to
5 described in any item double record quality detecting methods.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In realization such as double record quality detecting methods described in any one of claim 1 to 5 when the computer program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811537633.0A CN109767335A (en) | 2018-12-15 | 2018-12-15 | Double record quality detecting methods, device, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811537633.0A CN109767335A (en) | 2018-12-15 | 2018-12-15 | Double record quality detecting methods, device, computer equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109767335A true CN109767335A (en) | 2019-05-17 |
Family
ID=66450657
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811537633.0A Pending CN109767335A (en) | 2018-12-15 | 2018-12-15 | Double record quality detecting methods, device, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109767335A (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110266981A (en) * | 2019-06-17 | 2019-09-20 | 深圳壹账通智能科技有限公司 | Method, apparatus, computer equipment and the storage medium of video record |
CN110364183A (en) * | 2019-07-09 | 2019-10-22 | 深圳壹账通智能科技有限公司 | Method, apparatus, computer equipment and the storage medium of voice quality inspection |
CN111324713A (en) * | 2020-02-18 | 2020-06-23 | 腾讯科技(深圳)有限公司 | Automatic replying method and device for conversation, storage medium and computer equipment |
CN111401826A (en) * | 2020-02-14 | 2020-07-10 | 平安科技(深圳)有限公司 | Double-recording method and device for signing electronic contract, computer equipment and storage medium |
CN111461716A (en) * | 2020-04-02 | 2020-07-28 | 上海诺融信息科技有限公司 | Double recording system |
CN111723204A (en) * | 2020-06-15 | 2020-09-29 | 龙马智芯(珠海横琴)科技有限公司 | Method and device for correcting voice quality inspection area, correction equipment and storage medium |
CN111881825A (en) * | 2020-07-28 | 2020-11-03 | 深圳市点通数据有限公司 | Interactive text recognition method and system based on multi-perception data |
CN111931484A (en) * | 2020-07-31 | 2020-11-13 | 于梦丽 | Data transmission method based on big data |
CN112200537A (en) * | 2020-09-30 | 2021-01-08 | 中国银行股份有限公司 | Product confirmation method and system |
CN112699348A (en) * | 2020-12-25 | 2021-04-23 | 中国平安人寿保险股份有限公司 | Method and device for verifying nuclear body information, computer equipment and storage medium |
CN112818104A (en) * | 2021-02-05 | 2021-05-18 | 广州佰锐网络科技有限公司 | Intelligent interactive question and answer based account opening method and related system |
CN112884375A (en) * | 2021-03-29 | 2021-06-01 | 中信银行股份有限公司 | Application method and device based on double recording system, electronic equipment and medium |
CN112989021A (en) * | 2021-03-05 | 2021-06-18 | 上海明略人工智能(集团)有限公司 | Method, device and equipment for judging violation of counselor |
CN113035201A (en) * | 2021-03-16 | 2021-06-25 | 广州佰锐网络科技有限公司 | Financial service quality inspection method and system |
CN113095204A (en) * | 2021-04-07 | 2021-07-09 | 中国工商银行股份有限公司 | Double-recording data quality inspection method, device and system |
CN113111759A (en) * | 2021-04-07 | 2021-07-13 | 中国工商银行股份有限公司 | Customer confirmation detection method and device in double-record data quality inspection |
CN113223533A (en) * | 2021-05-31 | 2021-08-06 | 平安科技(深圳)有限公司 | Voice communication method and device, computer equipment and storage medium |
CN113255361A (en) * | 2021-05-19 | 2021-08-13 | 平安科技(深圳)有限公司 | Automatic detection method, device, equipment and storage medium of voice content |
CN113343824A (en) * | 2021-05-31 | 2021-09-03 | 平安银行股份有限公司 | Double-recording quality inspection method, device, equipment and medium |
WO2021196830A1 (en) * | 2020-03-30 | 2021-10-07 | 深圳壹账通智能科技有限公司 | Smart double-recording method and device, and storage medium |
CN115189911A (en) * | 2022-05-30 | 2022-10-14 | 平安科技(深圳)有限公司 | Generation method, device and equipment of surface label file and storage medium |
-
2018
- 2018-12-15 CN CN201811537633.0A patent/CN109767335A/en active Pending
Cited By (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110266981A (en) * | 2019-06-17 | 2019-09-20 | 深圳壹账通智能科技有限公司 | Method, apparatus, computer equipment and the storage medium of video record |
CN110364183A (en) * | 2019-07-09 | 2019-10-22 | 深圳壹账通智能科技有限公司 | Method, apparatus, computer equipment and the storage medium of voice quality inspection |
WO2021004128A1 (en) * | 2019-07-09 | 2021-01-14 | 深圳壹账通智能科技有限公司 | Voice quality control method and device, computer apparatus, and storage medium |
CN111401826A (en) * | 2020-02-14 | 2020-07-10 | 平安科技(深圳)有限公司 | Double-recording method and device for signing electronic contract, computer equipment and storage medium |
CN111324713A (en) * | 2020-02-18 | 2020-06-23 | 腾讯科技(深圳)有限公司 | Automatic replying method and device for conversation, storage medium and computer equipment |
CN111324713B (en) * | 2020-02-18 | 2022-03-04 | 腾讯科技(深圳)有限公司 | Automatic replying method and device for conversation, storage medium and computer equipment |
WO2021196830A1 (en) * | 2020-03-30 | 2021-10-07 | 深圳壹账通智能科技有限公司 | Smart double-recording method and device, and storage medium |
CN111461716A (en) * | 2020-04-02 | 2020-07-28 | 上海诺融信息科技有限公司 | Double recording system |
CN111723204A (en) * | 2020-06-15 | 2020-09-29 | 龙马智芯(珠海横琴)科技有限公司 | Method and device for correcting voice quality inspection area, correction equipment and storage medium |
CN111881825A (en) * | 2020-07-28 | 2020-11-03 | 深圳市点通数据有限公司 | Interactive text recognition method and system based on multi-perception data |
CN111881825B (en) * | 2020-07-28 | 2023-10-17 | 深圳市点通数据有限公司 | Interactive text recognition method and system based on multi-perception data |
CN111931484B (en) * | 2020-07-31 | 2022-02-25 | 贵州多彩宝互联网服务有限公司 | Data transmission method based on big data |
CN111931484A (en) * | 2020-07-31 | 2020-11-13 | 于梦丽 | Data transmission method based on big data |
CN112200537B (en) * | 2020-09-30 | 2024-02-02 | 中国银行股份有限公司 | Product confirmation method and system |
CN112200537A (en) * | 2020-09-30 | 2021-01-08 | 中国银行股份有限公司 | Product confirmation method and system |
CN112699348A (en) * | 2020-12-25 | 2021-04-23 | 中国平安人寿保险股份有限公司 | Method and device for verifying nuclear body information, computer equipment and storage medium |
CN112818104A (en) * | 2021-02-05 | 2021-05-18 | 广州佰锐网络科技有限公司 | Intelligent interactive question and answer based account opening method and related system |
CN112989021A (en) * | 2021-03-05 | 2021-06-18 | 上海明略人工智能(集团)有限公司 | Method, device and equipment for judging violation of counselor |
CN112989021B (en) * | 2021-03-05 | 2024-02-09 | 上海明略人工智能(集团)有限公司 | Method, device and equipment for advisor behavior violation determination |
CN113035201A (en) * | 2021-03-16 | 2021-06-25 | 广州佰锐网络科技有限公司 | Financial service quality inspection method and system |
CN112884375A (en) * | 2021-03-29 | 2021-06-01 | 中信银行股份有限公司 | Application method and device based on double recording system, electronic equipment and medium |
CN113111759A (en) * | 2021-04-07 | 2021-07-13 | 中国工商银行股份有限公司 | Customer confirmation detection method and device in double-record data quality inspection |
CN113095204A (en) * | 2021-04-07 | 2021-07-09 | 中国工商银行股份有限公司 | Double-recording data quality inspection method, device and system |
CN113255361A (en) * | 2021-05-19 | 2021-08-13 | 平安科技(深圳)有限公司 | Automatic detection method, device, equipment and storage medium of voice content |
CN113255361B (en) * | 2021-05-19 | 2023-12-22 | 平安科技(深圳)有限公司 | Automatic voice content detection method, device, equipment and storage medium |
CN113343824A (en) * | 2021-05-31 | 2021-09-03 | 平安银行股份有限公司 | Double-recording quality inspection method, device, equipment and medium |
CN113223533B (en) * | 2021-05-31 | 2023-06-06 | 平安科技(深圳)有限公司 | Voice call method, device, computer equipment and storage medium |
CN113223533A (en) * | 2021-05-31 | 2021-08-06 | 平安科技(深圳)有限公司 | Voice communication method and device, computer equipment and storage medium |
CN115189911A (en) * | 2022-05-30 | 2022-10-14 | 平安科技(深圳)有限公司 | Generation method, device and equipment of surface label file and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109767335A (en) | Double record quality detecting methods, device, computer equipment and storage medium | |
TWI773180B (en) | Computing system and method for calculating authenticity of human user | |
US10013972B2 (en) | System and method for identifying speakers | |
US11907274B2 (en) | Hyper-graph learner for natural language comprehension | |
WO2022095380A1 (en) | Ai-based virtual interaction model generation method and apparatus, computer device and storage medium | |
CN110378562A (en) | Voice quality detecting method, device, computer equipment and storage medium | |
CN112069484A (en) | Multi-mode interactive information acquisition method and system | |
CN109767321A (en) | Question answering process optimization method, device, computer equipment and storage medium | |
CN106919661A (en) | A kind of affective style recognition methods and relevant apparatus | |
CN106611447A (en) | Work attendance method and apparatus | |
Ottl et al. | Group-level speech emotion recognition utilising deep spectrum features | |
CN109919006A (en) | Expression detection method, device, electronic equipment and storage medium | |
CN109766419A (en) | Products Show method, apparatus, equipment and storage medium based on speech analysis | |
CN109887524A (en) | A kind of singing marking method, device, computer equipment and storage medium | |
US20210398135A1 (en) | Data processing and transaction decisioning system | |
Firc | Applicability of Deepfakes in the Field of Cyber Security | |
CN113469002A (en) | Identity recognition method based on block chain mutual authentication, biological multi-feature recognition and multi-source data fusion | |
CN113946663A (en) | Statement generation method, question asking method, device and readable medium | |
CN116308735A (en) | Financial data prediction method, device, electronic equipment and storage medium | |
CN116488850A (en) | Authenticity verification method and device | |
Orujova et al. | Personalizing Human-Robot Dialogue Interactions using Face and Name Recognition | |
Mokgonyane | Development of a text-independent automatic speaker recognition system | |
CN116863921A (en) | Speech recognition model training method, speech recognition method, device and equipment | |
CN113822326A (en) | Model evaluation method, model evaluation device, computer equipment and storage medium | |
CN110164434A (en) | The interview measures and procedures for the examination and approval and device based on speech recognition |
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 |