CN109472206A - Methods of risk assessment, device, equipment and medium based on micro- expression - Google Patents
Methods of risk assessment, device, equipment and medium based on micro- expression Download PDFInfo
- Publication number
- CN109472206A CN109472206A CN201811182176.8A CN201811182176A CN109472206A CN 109472206 A CN109472206 A CN 109472206A CN 201811182176 A CN201811182176 A CN 201811182176A CN 109472206 A CN109472206 A CN 109472206A
- Authority
- CN
- China
- Prior art keywords
- micro
- feature
- expression
- assessed
- characteristic
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/174—Facial expression recognition
- G06V40/176—Dynamic expression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/38—Payment protocols; Details thereof
- G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
- G06Q20/401—Transaction verification
- G06Q20/4016—Transaction verification involving fraud or risk level assessment in transaction processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Abstract
The invention discloses a kind of methods of risk assessment based on micro- expression, device, equipment and media.The described method includes: obtaining the first video data of object to be assessed;Micro- Expression Recognition is carried out to first video data using micro- Expression Recognition model, obtains first micro- expression data of object to be assessed;According to first micro- expression data, the micro- expression base-line data of face of object to be assessed is established;Obtain the second video data of object to be assessed;Micro- Expression Recognition is carried out to the second video data using micro- Expression Recognition model, obtains second micro- expression data of object to be assessed;Using the regime values range of every kind of index feature in the micro- expression base-line data of face, risk assessment is carried out to the characteristic to be identified of every kind of index feature in second micro- expression data, obtains the risk evaluation result of object to be assessed.Technical solution of the present invention realizes automatic progress risk assessment, and can be improved Credit Risk Assessment accuracy rate, reduces credit risk.
Description
Technical field
The present invention relates to field of computer technology more particularly to a kind of methods of risk assessment based on micro- expression, device, set
Standby and medium.
Background technique
As software systems are gradually applied to financial field, relatively broad answer especially has been obtained in loan financial field
With.Common loan financial system, such as credit system can be realized and fill out single process to loan, signing process, approval process, put
The electronization and automation of money process etc..
However, in the air control in credit process exchanged by aspectant between the careful people of letter and creditor at present,
Content of the people according to exchange and the observation subjective judgement to creditor and assessment are examined by believing, believes that examining people may be because attention
In the case where not concentrating or not knowing much have less understanding to the facial expression of people, so that some subtle expression shape changes of creditor are had ignored,
The difference for especially ignoring individual is to be difficult accurate discrimination risk exception only by subjective judgement, not especially for experience
The letter of foot examines people, is often difficult to identify potential high risk client, leads to higher credit risk.
Summary of the invention
The embodiment of the present invention provides a kind of methods of risk assessment based on micro- expression, device, equipment and medium, to solve mesh
Preceding Credit Risk Assessment accuracy rate is not high, leads to the problem that credit risk is high.
A kind of methods of risk assessment based on micro- expression, comprising:
Obtain the first video data of object to be assessed, wherein first video data returns for the object to be assessed
Answer the video data of preset underlying issue;
Micro- Expression Recognition is carried out to first video data using preset micro- Expression Recognition model, is obtained described to be evaluated
Estimate first micro- expression data of object, wherein preset micro- Expression Recognition model is identified from first video data
The characteristic of preset index feature and every kind of index feature, first micro- expression data includes every kind of index
The characteristic of feature;
According to the characteristic of every kind of index feature in described first micro- expression data, the object to be assessed is established
The micro- expression base-line data of face, wherein the micro- expression base-line data of face includes the normal number of every kind of index feature
It is worth range;
Obtain the second video data of the object to be assessed, wherein second video data is described to be assessed right
Video data as answering preset evaluation problem;
Micro- Expression Recognition is carried out to second video data using micro- Expression Recognition model, is obtained described to be assessed
The micro- expression data of the second of object, wherein preset micro- Expression Recognition model identifies institute from second video data
The characteristic to be identified of index feature and every kind of index feature is stated, second micro- expression data includes every kind of institute
State the characteristic to be identified of index feature;
Using the regime values range of every kind of index feature in the micro- expression base-line data of the face, to described second
The characteristic to be identified of every kind of index feature carries out risk assessment in micro- expression data, obtains the object to be assessed
Risk evaluation result.
A kind of risk assessment device based on micro- expression, comprising:
First obtains module, for obtaining the first video data of object to be assessed, wherein first video data is
The object to be assessed answers the video data of preset underlying issue;
First identification module, for carrying out micro- expression to first video data using preset micro- Expression Recognition model
Identification, obtains first micro- expression data of the object to be assessed, wherein preset micro- Expression Recognition model is from described the
The characteristic of preset index feature and every kind of index feature, first micro- expression are identified in one video data
Data include the characteristic of every kind of index feature;
Baseline establishes module, for the characteristic according to every kind of index feature in described first micro- expression data,
Establish the micro- expression base-line data of face of the object to be assessed, wherein the micro- expression base-line data of face includes every kind of institute
State the regime values range of index feature;
Second obtains module, for obtaining the second video data of the object to be assessed, wherein second video counts
According to the video data for answering preset evaluation problem for the object to be assessed;
Second identification module, for carrying out micro- expression knowledge to second video data using micro- Expression Recognition model
Not, second micro- expression data of the object to be assessed is obtained, wherein preset micro- Expression Recognition model is from described second
The characteristic to be identified of the index feature and every kind of index feature, second micro- table are identified in video data
Feelings data include the characteristic to be identified of every kind of index feature;
Risk evaluation module, for using the normal number of every kind of index feature in the micro- expression base-line data of the face
It is worth range, risk assessment is carried out to the characteristic to be identified of every kind of index feature in described second micro- expression data, is obtained
To the risk evaluation result of the object to be assessed.
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 that the above-mentioned risk based on micro- expression is commented when executing the computer program
The step of estimating method.
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 methods of risk assessment based on micro- expression when being executed by processor.
In the above-mentioned methods of risk assessment based on micro- expression, device, equipment and medium, answered by acquiring object to be assessed
First video data of preset underlying issue, and micro- table is carried out to the first video data using preset micro- Expression Recognition model
Feelings identification, obtains first micro- expression data comprising object to be assessed in the characteristic of every kind of index feature, and then basis should
First micro- expression data establishes the micro- expression base-line data of face of object to be assessed, and the micro- expression base-line data of the face includes every kind
The regime values range of index feature realizes and establishes a set of micro- expression baseline number of independent face for each object to be assessed
According to;It is preset by obtaining object answer to be assessed on the basis of establishing facial micro- expression base-line data to object to be assessed
Second video data of evaluation problem, and micro- expression knowledge is carried out to the second video data using identical micro- Expression Recognition model
, do not obtain comprising object to be assessed the characteristic to be identified of every kind of index feature second micro- expression data, then with to
On the basis of the micro- expression base-line data of face for assessing object, to the feature to be identified of every kind of index feature in second micro- expression data
Data carry out risk assessment, obtain the risk evaluation result of object to be assessed, on the one hand, pass through preset micro- Expression Recognition model
Carry out micro- Expression Recognition, and various index features relevant to micro- expression be set, according to the characteristic of index feature compare into
Row risk assessment realizes and realizes risk assessment to object to be assessed automatically based on micro- expression, and artificial subjective evaluation is avoided to believe
The low problem of accuracy rate caused by risk is borrowed, Credit Risk Assessment accuracy rate is improved;On the other hand, by being each to be assessed right
As establishing the micro- expression base-line data of independent face, enable when carrying out risk assessment using the micro- expression base-line data of face
Individual difference is distinguished, to further effectively improve Credit Risk Assessment accuracy rate, reduces credit risk.
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 the methods of risk assessment based on micro- expression in one embodiment of the invention;
Fig. 2 is a flow chart of the methods of risk assessment based on micro- expression in one embodiment of the invention;
Fig. 3 is a flow chart of step S2 in methods of risk assessment based on micro- expression in one embodiment of the invention;
Fig. 4 is a flow chart of step S3 in methods of risk assessment based on micro- expression in one embodiment of the invention;
Fig. 5 is a flow chart of step S6 in risk assessment based on micro- expression in one embodiment of the invention;
Fig. 6 is a schematic diagram of the risk assessment device in one embodiment of the invention based on micro- expression;
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.
Methods of risk assessment provided by the present application based on micro- expression, can be applicable in application environment as shown in Figure 1, should
Application environment includes server-side and client, wherein is attached between server-side and client by network, which can be with
It is cable network or wireless network, client is specifically including but not limited to various personal computers, laptop, intelligent hand
Machine, tablet computer and portable wearable device, server-side can specifically be formed with independent server or multiple servers
Server cluster realize.The video data of collected object to be assessed is sent to server-side by client, server-side according to
The video data received carries out risk assessment.
In one embodiment, it as shown in Fig. 2, providing a kind of methods of risk assessment based on micro- expression, applies in this way
It is illustrated for server-side in Fig. 1, details are as follows:
S1: the first video data of object to be assessed is obtained, wherein the first video data is that object to be assessed answer is default
Underlying issue video data.
Specifically, object to be assessed is the user for needing to carry out risk assessment, when need to object to be assessed carry out risk
When assessment, client issues underlying issue acquisition request to server-side, and preset underlying issue is sent to client by server-side,
Client shows underlying issue to object to be assessed, and acquires the video counts to be assessed to answer underlying issue by acquisition device
According to rear, which is sent to server-side, which specifically can be video file.Server-side receives client hair
The video data sent, the first video data as object to be assessed.
Wherein, underlying issue can be personal information problem, for example, age, gender, ID card No., phone number, with
And home address etc..
S2: micro- Expression Recognition is carried out to the first video data using preset micro- Expression Recognition model, it is to be assessed right to obtain
The micro- expression data of the first of elephant, wherein preset micro- Expression Recognition model identifies that preset index is special from the first video data
The characteristic of sign and every kind of index feature, first micro- expression data includes the characteristic of every kind of index feature.
Specifically, server-side is built collected video data in step S1 and is inputted in preset micro- Expression Recognition model, should
Micro- Expression Recognition model extracts video frame from video data, and carries out micro- Expression Recognition to the facial image in video frame, mentions
Take the characteristic of various index features of the object to be identified under normal micro- emotional state.
Wherein, micro- Expression Recognition model specifically can be the neural network model based on deep learning, and preset index is special
Sign includes but is not limited to AU (Action Unit) index, headwork index, eye motion index etc., the feature of index feature
Data specifically can be the times or frequency that the index feature occurs within a preset period of time.
For example, preset time period is 2 minutes, index feature includes eye motion feature and headwork feature, and eye
Portion's motion characteristic includes blink feature, and headwork feature includes that head twists feature to the left and twists feature to the right, then passes through
First micro- expression data that micro- Expression Recognition model identifies includes: that the number of winks in 2 minutes is 12 times, and head is to the left
The number of twisting is 6 times, and the number that head twists to the right is 4 times.
Further, if not identifying the characteristic of some index feature, this is referred to according to group's normal distribution
The characteristic of mark feature is set as preset default data.
For example, head per minute twists 1 time most people to the left under normal circumstances, therefore head per minute is twisted to the left
Twist the default data of feature for 1 time to the left as head.If not identifying that head twists feature to the left to object to be assessed
Characteristic, then head is twisted to the left feature characteristic be set as default head per minute twist to the left 1 time.
S3: it according to the characteristic of every kind of index feature in first micro- expression data of object to be assessed, establishes to be assessed
The micro- expression base-line data of face of object, wherein the micro- expression base-line data of face includes the regime values model of every kind of index feature
It encloses.
Specifically, first micro- expression data includes the characteristic of every kind of index feature, with the feature of every kind of index feature
Data are core according to the progress range extension of preset ratio, the regime values range of every kind of index feature are obtained, as to be evaluated
Estimate the micro- expression base-line data of face of object.
For example, if the characteristic of blink feature is blink 12 times per minute, with 12 times per minute for core, according to pre-
If ratio carry out range extension, it is assumed that preset ratio be 50%, then will blink per minute 6 times to 18 times as blink features
Regime values range.
It should be noted that the micro- expression base-line data of face of object to be assessed refers to object to be assessed in normal micro- expression
Micro- expression data under state, in the present embodiment, preset underlying issue are the problem of clearly knowing correct option, to
Micro- expression for showing when answering these underlying issues of assessment object is considered as normal micro- expression, therefore, default to
It assesses when object answers underlying issue and is obtained in normal micro- emotional state, and according to the characteristic under normal micro- emotional state
To base-line data can accurately express normal micro- expression of state to be assessed, and carry out wind in this, as to object to be assessed
The foundation nearly assessed has the specific aim of individual so that each base-line data is directed to specific object to be assessed, distinguishes individual
Difference, so as to improve the accuracy of the risk assessment to object to be assessed.
S4: the second video data of object to be assessed is obtained, wherein the second video data is that object to be assessed answer is default
Evaluation problem video data.
Preset evaluation problem is the tender subject for credit risk setting, by object to be assessed to evaluation problem
Answer can judge the credit risk of the object to be assessed, for example, evaluation problem specifically can be loan user, personal income,
And repay wish etc..
Specifically, client issues evaluation problem acquisition request to server-side, and server-side sends preset evaluation problem
To client, client shows evaluation problem to object to be assessed, and acquires to be assessed assess answer by acquisition device and ask
After the video data of topic, which is sent to server-side, which specifically can be video file.Service termination
Receive the video data that client is sent, the second video data as object to be assessed.
S5: micro- Expression Recognition is carried out to the second video data using preset micro- Expression Recognition model, it is to be assessed right to obtain
The micro- expression data of the second of elephant, wherein preset micro- Expression Recognition model identifies that preset index is special from the second video data
The characteristic to be identified of sign and every kind of index feature, second micro- expression data includes the spy to be identified of every kind of index feature
Levy data.
Specifically, server-side uses the second video obtained with micro- Expression Recognition model identical in step S2 to step S4
Data carry out micro- Expression Recognition, identification process with it is micro- to the progress of the first video data using micro- Expression Recognition model in step S2
The identification process of Expression Recognition is consistent, and to avoid repeating, details are not described herein again.
Second micro- expression data that server-side obtains after micro- Expression Recognition is that object to be assessed is answering evaluation problem
When the characteristic of the index feature of micro- expression that shows, for this feature data by as characteristic to be identified, server-side can
Whether with normal according to the micro- emotional state for the analytical judgment object to be assessed for treating identification feature data, and then determination is to be assessed
Object is with the presence or absence of credit risk and the degree of credit risk.
S6: right using the regime values range of every kind of index feature in the micro- expression base-line data of face of object to be assessed
The characteristic to be identified of every kind of index feature carries out risk assessment in second micro- expression data, obtains the risk of object to be assessed
Assessment result.
Specifically, server-side is according to the micro- expression base-line data of face of the obtained object to be assessed of step S3, to step S5
The micro- expression data of second obtained carries out risk assessment, obtains the risk evaluation result of object to be assessed.
It should be noted that risk evaluation result specifically can be the numerical value of quantization, and the bigger expression risk of numerical value is more
Height, risk evaluation result can also be preset evaluation grade, for example, the first estate, second grade etc., and higher grade table
Show that risk is bigger, risk evaluation result can also using other definition modes, specifically can according to the needs of practical application into
Row setting, herein with no restrictions.
In one embodiment, the process of risk assessment can be for every kind of index feature, compare the index feature
Characteristic to be identified and the index feature normal data range between deviation, it is assumed that preset index feature is M,
M deviation is then obtained, server-side is ranked up M deviation according to sequence from big to small, and chooses the preceding N of sequence
A deviation calculates the average value of N number of deviation, using the average value as the risk evaluation result of object to be assessed,
In, M and N are positive integer, and M is more than or equal to N.
In the present embodiment, the first video data of preset underlying issue is answered by acquiring object to be assessed, and use
Preset micro- Expression Recognition model carries out micro- Expression Recognition to the first video data, obtains comprising object to be assessed in every kind of index
The micro- expression data of the first of the characteristic of feature, the face for then establishing object to be assessed according to first micro- expression data are micro-
Expression base-line data, the micro- expression base-line data of the face include the regime values range of every kind of index feature, are realized for every
A object to be assessed establishes a set of micro- expression base-line data of independent face;The micro- expression baseline of face is being established to object to be assessed
On the basis of data, the second video data of preset evaluation problem is answered by obtaining object to be assessed, and using identical
Micro- Expression Recognition model carries out micro- Expression Recognition to the second video data, obtains including object to be assessed in every kind of index feature
The micro- expression data of the second of characteristic to be identified, it is right then on the basis of the micro- expression base-line data of face of object to be assessed
The characteristic to be identified of every kind of index feature carries out risk assessment in second micro- expression data, obtains the risk of object to be assessed
Assessment result, on the one hand, micro- Expression Recognition is carried out by preset micro- Expression Recognition model, and is arranged relevant to micro- expression each
Kind of index feature compares according to the characteristic of index feature and carries out risk assessment, realizes and is realized automatically based on micro- expression pair
The risk assessment of object to be assessed, the problem for avoiding accuracy rate caused by artificial subjective evaluation credit risk low improve credit wind
Danger assessment accuracy rate;On the other hand, by establishing the micro- expression base-line data of independent face for each object to be assessed, so that
Individual difference can be distinguished when carrying out risk assessment using the micro- expression base-line data of face, to further effectively improve credit wind
Danger assessment accuracy rate, reduces credit risk.
In one embodiment, preset micro- Expression Recognition model includes Face datection model, mood discrimination model, head appearance
State identification model, blink detection model and iris edge detection model, preset index feature include muscle movement feature, head
Posture feature, blink feature and eye movement variation characteristic.
Specifically, the Face datection model human face region in video data in video frame images for identification;Mood differentiates
Model be used for human face region carry out muscle movement feature identification, muscle movement feature, that is, AU index, for example, interior eyebrow raise up,
The corners of the mouth raises up, nose crease etc.;Head pose identification model for identification in human face region head in the inclined of different preset directions
Angle, i.e. head pose feature are moved, for example, twisting to the left, to the right twisting etc.;Blink detection model is used for from continuous time period
Frequency of wink is detected in the human face region of video frame images, i.e. blink feature;Iris edge detection model face area for identification
The case where eye motion changes in domain, i.e. eye movement variation characteristic, for example, turning left, rotating backward.
Preferably, in the present embodiment, the characteristic of every kind of index feature can be the frequency of index feature appearance.
Further, as shown in figure 3, in step s 2, using preset micro- Expression Recognition model to the first video data
Micro- Expression Recognition is carried out, first micro- expression data of object to be assessed is obtained, specifically comprises the following steps:
S21: according to preset extracting mode, the video frame images of the first preset quantity are extracted from the first video data.
Specifically, preset extracting mode can be from the video frame images that the first video data includes, at interval of pre-
Fixed interval frame number extracts a video frame images, obtains the video frame images of preset quantity.
For example, including 1000 video frame images in the first video data, a video frame figure is extracted according at interval of 5 frames
The extracting mode of picture extracts 200 video frame images from first video data.
S22: using face detection model, carries out Face datection to video frame images, extracts the face in video frame images
Picture.
Specifically, preset Face datection model can use cascade connection type convolutional neural networks (CascadeCNN) model,
It realizes fast face detection, has 6 convolutional neural networks, 3 convolutional neural networks in the cascade structure of CascadeCNN model
For face and two non-face classification, in addition 3 convolutional neural networks are corrected for the frame of human face region, to obtain people
Face picture.
It should be noted that when carrying out face picture extraction using face detection model, if including in video frame images
Multiple face pictures are then extracted and occupy the maximum human face region of image area as the face picture extracted, that is, require to be evaluated
Estimate the default nucleus that object is in video acquisition in answer to a question.
S23: face picture is inputted into mood discrimination model, micro- Expression Recognition is carried out, obtains every kind of flesh of object to be assessed
The characteristic of meat motion characteristic.
Specifically, muscle movement feature, that is, AU feature, the basic muscle movement unit including face, such as in eyebrow raise up,
The corners of the mouth raises up, nose crease etc..
In mood discrimination model, multiple muscle movement features are preset, by micro- Expression Recognition to face picture,
Identify the number that each muscle movement feature occurs within the first video data corresponding period, and according to the first video counts
According to the number that corresponding period and each muscle movement feature occur, the frequency of muscle movement feature appearance is calculated, and will
Characteristic of the frequency as the muscle movement feature.
Mood discrimination model can specifically use depth convolutional neural networks (DCNN) model and 80 layers of residual error network
(ResNet-80) model.
It is identified using position coordinates of the DCNN model to human face characteristic point, wherein human face characteristic point includes but unlimited
In left eye, right eye, nose, the left corners of the mouth, the right corners of the mouth, and the recognition result of DCNN model is input in ResNet-80 model.
In ResNet-80 model, according to the position coordinates of the human face characteristic point of input, the people is extracted from face picture
The characteristic value of face characteristic point, and carry out similarity with the Standard Eigenvalue of preset each AU feature and compare, and according to similarity
Comparison result judges whether the human face characteristic point meets the requirement of each AU feature, confirms that the AU feature occurs one if meeting
It is secondary, the total degree that each AU feature occurs is counted after the completion of comparing to each human face characteristic point, and true according to the total degree
The characteristic of fixed every kind of muscle movement feature.
S24: face picture is inputted into head gesture recognition model, head pose identification is carried out, obtains object to be assessed
The characteristic of every kind of head pose feature.
Specifically, 10 layers of convolutional neural networks model can be used in head pose identification model, pass through different convolution kernels
It carries out convolution operation and obtains the characteristic pattern of head feature, meanwhile, in order to avoid linear model ability to express is inadequate, further use
Nonlinear function carries out non-linearization operation to the characteristic pattern obtained after convolution, prevents over-fitting, the nonlinear function packet of use
Include but be not limited to sigmoid, tanh and ReLU etc..
Preset head pose feature may include the head twisting feature of multiple preset directions, wherein preset direction packet
Include but be not limited to top to bottom, left and right, front and rear etc..
Head pose identification model, which passes through, judges that the head of each preset direction twists whether angle is more than the preset direction
Angle threshold, determine and twist number on the head of the preset direction, if the head twisting angle of the preset direction is more than that this is pre-
The angle threshold of set direction, it is determined that twisting is primary on the preset direction head, and counts the first video data corresponding time
The total degree that each head twisting feature occurs in section, and twisted according to the first video data corresponding period and each head
The total degree that feature occurs calculates the frequency that each head twisting feature occurs, and twists feature for the frequency as the head
Characteristic.
For example, in one embodiment, carrying out head pose identification to face picture by head pose identification model
Afterwards, obtaining the frequency that object header to be assessed twists to the left is 6 beats/min, and the frequency that head twists to the right is 8 beats/min
Characteristic.
S25: face picture is inputted into blink detection model, number of winks detection is carried out, obtains frequency of wink, and this is blinked
Characteristic of the eye frequency as the blink feature of object to be assessed.
Specifically, blink detection model can judge people using logistic regression (Logistic Regression, LR) model
Whether object to be assessed blinks in face picture.Wherein, Logic Regression Models are one of machine learning disaggregated models,
It is two disaggregated models.
Every face picture is inputted into preset blink detection model, obtains hair in the first video data corresponding period
The quantity of the face picture of raw blink, and using the quotient of the quantity and the period as frequency of wink, that is, the feature for feature of blinking
Data.
S26: inputting iris edge detection model for face picture, carries out eye movement variation detection, obtains object to be assessed
The characteristic of every kind of eye movement variation characteristic.
Specifically, iris edge detection model determines eye center position in eye by detecting the position of eye iris edge
The variation track in socket of the eye region, thus the case where obtaining eye movement variation, i.e., the characteristic of preset eye movement variation characteristic, eye movement becomes
Changing feature includes but is not limited to that eyes are moved to the left, and eyes move right, and eyes move up, and is directed one's eyes downward mobile etc..
In iris edge detection model, the eyeball centre coordinate point in face picture is obtained, with eyeball centre coordinate point
Centered on, eye orbit areas is cut out with preset radius, and the position of iris edge point is detected in eye areas, according to iris
The position of marginal point determines the enclosed region that iris edge point surrounds, and the center of the enclosed region is determined as eye center
Position, real-time tracing eye center position identify the mobile side of eyes according to the variation track in the variation track of eye orbit areas
To, the statistics eyes number mobile to each preset direction, and counted according to the number and the first video data corresponding period
Calculate the eyes frequency mobile to each preset direction, the characteristic of as every kind eye movement variation characteristic.
It should be noted that there is no inevitable successively execution suitable between step S23, step S24, step S25 and step S26
Sequence can be the relationship executed side by side, herein with no restrictions.
S27: by the characteristic of every kind of muscle movement feature, the characteristic of every kind of head pose feature, blink feature
Characteristic and every kind of eye variation characteristic first micro- expression data of the characteristic as object to be assessed.
Specifically, by the characteristic of every kind of obtained muscle movement feature of step S23, every kind of head that step S24 is obtained
The characteristic of portion's posture feature, every kind of eye that the characteristic and step S26 for the blink feature that step S25 is obtained obtain
The characteristic of portion's variation characteristic is combined into first micro- expression data of object to be assessed.
In the present embodiment, according to preset extracting mode, the video frame figure of preset quantity is extracted from the first video data
Picture includes using micro- Expression Recognition model respectively later by the face picture in Face datection model extraction video frame images
Mood discrimination model, head pose identification model, blink detection model and iris edge detection model, to face picture carry out
Feature identification obtains the characteristic of every kind of different index features, and is combined into first micro- expression data of object to be assessed, real
Show from multiple and different angles and dimension and various dimensions excavation is carried out to the micro- expression of face, it is to be assessed right more accurately to embody
Micro- expression feature of elephant establishes baseline according to first micro- expression data and provides accurate data basis, to be conducive to be subsequent
More scientific and reasonable obtains more accurate risk evaluation result.
It is understood that carrying out micro- Expression Recognition to the second video data using micro- Expression Recognition model, obtain to
When assessing second micro- expression data of object, also need to obtain second micro- expression using recognition methods same with the above-mentioned embodiment
Data, so that risk assessment can be based on identical data collecting standard, it is ensured that the accuracy of assessment.
In one embodiment, as shown in figure 4, in step s3, according to every in first micro- expression data of object to be assessed
The characteristic of kind index feature is established the micro- expression base-line data of face of object to be assessed, is specifically comprised the following steps:
S31: the corresponding default minimum scale coefficient of every kind of index feature and default maximum in first micro- expression data are obtained
Proportionality coefficient.
Specifically, server-side is that default minimum scale coefficient and default maximum ratio system is arranged in every kind of index feature in advance
Number.Default minimum scale coefficient and default maximum ratio coefficient are used for the regime values range of regulating index feature.
For example, corresponding default minimum scale can be 0.5, corresponding pre- to blink this index feature of feature
If maximum ratio can be 1.5.
S32: be directed to every kind of index feature, calculate the index feature characteristic it is corresponding with the index feature preset most
The characteristic of the first product and the index feature between small scale coefficient default high specific corresponding with the index feature
The second product between example coefficient.
Specifically, it is assumed that the value of the characteristic of some index feature is w, the corresponding default minimum scale of the index feature
Coefficient is a, and presetting maximum ratio coefficient is b, then the value of the first product is w*a, and the value of the second product is w*b.
Continue by taking the blink feature in step S31 as an example, if the characteristic of blink feature is 12 beats/min, i.e. w=
12, then the value of the first product is 12*0.5=6, and the value of the second product is 12*1.5=18.
S33: by the value range between corresponding the second product of first sum of products of every kind of index feature, it is determined as every kind of finger
The regime values range for marking feature obtains the micro- expression base-line data of face of object to be assessed.
Specifically, first product that will be greater than or equal to and be less than or equal to the second product numerical value be determined as index feature just
Constant value range.
Continue by taking the blink feature in step S31 as an example, regime values range is [6,18].
In the present embodiment, according to the corresponding default minimum scale coefficient of every kind of index feature and default maximum ratio coefficient,
The characteristic of every kind of index feature in first micro- expression data is rationally expanded, the normal number of every kind of index feature is obtained
The scientific and reasonable determining micro- expression baseline of face is realized as the micro- expression base-line data of face of object to be assessed according to range
Data, to be conducive to carry out accurate risk assessment to object to be assessed according to the micro- expression base-line data of the face.
In one embodiment, as shown in figure 5, in step s 6, using the micro- expression base-line data of face of object to be assessed
In every kind of index feature regime values range, to the characteristic to be identified of every kind of index feature in second micro- expression data into
Row risk assessment obtains the risk evaluation result of object to be assessed, specifically comprises the following steps:
S61: it is directed to every kind of index feature, if the characteristic to be identified of the index feature meets the normal of the index feature
The requirement of numberical range then confirms the index feature devoid of risk of object to be assessed, otherwise, confirms the index of object to be assessed
There are risks for feature.
Specifically, to preset every kind of index feature, judge whether the characteristic to be identified of the index feature belongs to this
Within the scope of the regime values of index feature, if the characteristic to be identified of the index feature belongs to the regime values of the index feature
Range meets the requirement of the regime values range of the index feature, then confirm the index feature devoid of risk, if the index feature
Characteristic to be identified be not belonging to the regime values range of the index feature, that is, be unsatisfactory for the regime values model of the index feature
The requirement enclosed then confirms that there are risks for the index feature.
In one embodiment, the risk identification that a bit is arranged to each index feature in server-side accords with, and uses
Risk identification symbol mark index feature whether there is risk, if there are risks for index feature, by the risk of the index feature
Identifier is set as 1, otherwise, if risk is not present in index feature, sets 0 for the risk identification symbol of the index feature.
S62: the spy to be identified of the index feature is calculated if there are risks for the index feature for every kind of index feature
The ratio of regime values range of the data beyond the index feature is levied, and according to the index of ratio-dependent object to be assessed spy
The risk score of sign.
Specifically, the judging result that whether there is risk according to every kind of index feature that step S61 is obtained, if the index is special
There are risks for sign, then obtain the characteristic to be identified of the index feature, by the characteristic to be identified and the index feature
Regime values range is compared, and is calculated the characteristic to be identified beyond the difference of the regime values range and is accounted for the regime values
The ratio of range, the ratio including being greater than the regime values range maximum value, or less than the normal data stated range minimum
Ratio.
For example, continuing by taking the blink feature in step S31 as an example, regime values range is [6,18], if its is to be identified
Characteristic is 25, i.e., frequency of wink of the object to be assessed when answering evaluation problem is 25 beats/min, then the feature to be identified
Ratio of the data beyond the regime values range is the ratio greater than 18, i.e. (25-18)/18=39%;If its feature to be identified
Data are 2, i.e., frequency of wink of the object to be assessed when answering evaluation problem is 2 beats/min, then the characteristic to be identified is super
The ratio of the regime values range is the ratio less than 6, i.e. (6-2)/6=66.7% out.
Server-side presets the direct proportion linear functional relation X=λ * h+ δ of ratio h and risk score X, wherein λ and δ
It is preset parameter, and each index feature corresponds to independent parameter lambda and δ, i.e., different index features, corresponding ginseng
Number λ and δ can be identical or not identical, needs to be arranged with specific reference to practical application.
According to the corresponding direct proportion linear functional relation of each index feature, the to be identified of the index feature of risk will be present
Characteristic brings corresponding direct proportion linear functional relation into beyond the ratio of regime values range, this is calculated, and there are risks
Index feature risk score.
S63: to object to be assessed, there are the risk scores of the index feature of risk to be weighted, and it is to be assessed right to obtain
The risk class score of elephant.
Specifically, it each of is obtained according to step S62 there are the risk score of the index feature of risk, according to preset every
The Risk rated ratio of a index feature, is weighted, risk class score of the obtained result as object to be assessed.
S64: according to the corresponding relationship of preset risk class fraction range and risk class, the wind of object to be assessed is determined
Risk class fraction range where dangerous rating fraction, and by the corresponding risk class of risk class fraction range be determined as to
Assess the risk evaluation result of object.
Specifically, server-side presets the corresponding relationship of risk class fraction range and risk class, according to step S63
Obtained risk class score determines the risk class fraction range where the risk class score, and according to corresponding relationship, obtains
The corresponding risk class of risk class fraction range is taken, and using the risk class as the risk assessment knot of object to be assessed
Fruit.
In one embodiment, preset risk class includes slight risk, average risk, medium risk and serious wind
Four grades in danger, risk class fraction range of the risk class score less than 30 correspond to slight risk, and risk class score is greater than
Risk class fraction range equal to 30 and less than 60 corresponds to average risk, and risk class score is more than or equal to 60 and is less than
80 risk class fraction range corresponds to medium risk, and risk class fraction range of the risk class score more than or equal to 80 is corresponding
Serious risk.
For example, confirming that the risk evaluation result of object to be assessed is if the risk class score that step S63 is obtained is 70
Medium risk.
In the present embodiment, whether the normal of the index feature is met by the characteristic to be identified of judge index feature
The requirement of numberical range confirms that the index feature whether there is risk, and for there are the index features of risk, according to the index
The ratio of regime values range of the characteristic to be identified of feature beyond the index feature calculates the risk point of the index feature
Number, and to being weighted to obtain risk class score there are the risk score of the index feature of risk, according to the risk etc.
Grade score determines risk class, so that the risk evaluation result of object to be assessed is the index feature based on each different dimensions
Comprehensive descision is analyzed and quantifies to obtain, and accuracy rate is higher, and converts corresponding risk class for the score after quantization, so that knot
Fruit is more intuitive.
In one embodiment, in step S63, to object to be assessed, there are the progress of the risk score of the index feature of risk
Weighted calculation obtains the risk class score of object to be assessed, specifically comprises the following steps:
The risk class score of object to be assessed is calculated according to following formula (1):
Wherein, P is the risk class score of object to be assessed, kiFor i-th, there are the default power of the index feature of risk
Weight, xiFor i-th there are the risk score of the index feature of risk, n is the quantity of the index feature there are risk, and β is default
Adjustment parameter.
For example, the index feature of risk is that blink feature and eyes move up feature if it exists, and feature of blinking
Default weight is 0.6, and the default weight that eyes move up feature is 0.2, and preset adjustment parameter is 10, the wind for feature of blinking
Dangerous score is 55, and the risk score that eyes move up feature is 70, then the risk class point being calculated according to above-mentioned formula
Number is 55*0.6+70*0.2+10=57.
In the present embodiment, according to the risk class score for the object to be assessed that formula (1) is calculated, each deposit is combined
In the weight and risk score of the index feature of risk, risk class score accurately and is comprehensively reflected to be assessed right
Micro- emotional state of elephant, so that the accuracy rate of the risk evaluation result obtained according to risk class score is higher.
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 risk assessment device based on micro- expression is provided, it should the risk assessment based on micro- expression
Methods of risk assessment in device and above-described embodiment based on micro- expression corresponds.As shown in fig. 6, should the wind based on micro- expression
Danger assessment device includes that the first acquisition module 61, the first identification module 62, baseline establish module 63, second and obtain module 64, the
Two identification modules 65 and risk evaluation module 66.Detailed description are as follows for each functional module:
First obtain module 61, for obtaining the first video data of object to be assessed, wherein the first video data be to
Assessment object answers the video data of preset underlying issue;
First identification module 62, for carrying out micro- expression knowledge to the first video data using preset micro- Expression Recognition model
Not, first micro- expression data of object to be assessed is obtained, wherein preset micro- Expression Recognition model is known from the first video data
The characteristic of not preset index feature and every kind of index feature, first micro- expression data include every kind of index feature
Characteristic;
Baseline establishes module 63, for the characteristic according to every kind of index feature in first micro- expression data, establish to
Assess the micro- expression base-line data of face of object, wherein the micro- expression base-line data of face includes the normal number of every kind of index feature
It is worth range;
Second obtain module 64, for obtaining the second video data of object to be assessed, wherein the second video data be to
Assessment object answers the video data of preset evaluation problem;
Second identification module 65 is obtained for carrying out micro- Expression Recognition to the second video data using micro- Expression Recognition model
To second micro- expression data of object to be assessed, wherein preset micro- Expression Recognition model identifies in advance from the second video data
If index feature and every kind of index feature characteristic to be identified, second micro- expression data includes every kind of index feature
Characteristic to be identified;
Risk evaluation module 66, for using the regime values model of every kind of index feature in the micro- expression base-line data of face
It encloses, risk assessment is carried out to the characteristic to be identified of every kind of index feature in second micro- expression data, obtains object to be assessed
Risk evaluation result.
Further, preset micro- Expression Recognition model includes Face datection model, mood discrimination model, head pose knowledge
Other model, blink detection model and iris edge detection model, preset index feature include muscle movement feature, head pose
Feature, blink feature and eye movement variation characteristic, the first identification module 62 include:
Frame image zooming-out submodule, for extracting preset quantity from the first video data according to preset extracting mode
Video frame images;
Face datection submodule carries out Face datection to video frame images, extracts video for using face detection model
Face picture in frame image;
Mood differentiates submodule, for face picture to be inputted mood discrimination model, carries out micro- Expression Recognition, obtains to be evaluated
Estimate the characteristic of every kind of muscle movement feature of object;
Head pose identifies submodule, for face picture to be inputted head gesture recognition model, carries out head pose knowledge
Not, the characteristic of every kind of head pose feature of object to be assessed is obtained;
Blink detection submodule carries out number of winks detection, is blinked for face picture to be inputted blink detection model
Eye frequency, and using frequency of wink as the characteristic of the blink feature of object to be assessed;
Eye movement changes detection sub-module, for face picture to be inputted iris edge detection model, carries out eye movement variation inspection
It surveys, obtains the characteristic of every kind of eye movement variation characteristic of object to be assessed;
Micro- expression data generates submodule, for by the characteristic of every kind of muscle movement feature, every kind of head pose spy
The characteristic of the characteristic of sign, the characteristic for feature of blinking and every kind of eye variation characteristic is as first micro- expression number
According to.
Further, baseline establishes module 63 and includes:
Coefficient acquisition submodule, for obtaining the corresponding default minimum scale of every kind of index feature in first micro- expression data
Coefficient and default maximum ratio coefficient;
Data expansion submodule calculates the characteristic and the index of the index feature for being directed to every kind of index feature
The characteristic and the index feature of the first product and the index feature between the corresponding default minimum scale coefficient of feature
The second product between corresponding default maximum ratio coefficient;
Baseline determines submodule, for by the value model between corresponding the second product of first sum of products of every kind of index feature
It encloses, is determined as the regime values range of every kind of index feature, obtain the micro- expression base-line data of face of object to be assessed.
Further, risk evaluation module 66 includes:
Index risk determines submodule, for being directed to every kind of index feature, if the characteristic to be identified of the index feature
The requirement for meeting the regime values range of the index feature then confirms the index feature devoid of risk of object to be assessed, otherwise, really
Recognizing the index feature of object to be assessed, there are risks;
Risk score computational submodule, for being directed to every kind of index feature, if the index feature, there are risk, calculating should
The ratio of regime values range of the characteristic to be identified of index feature beyond the index feature, and waited for according to the ratio-dependent
Assess the risk score of the index feature of object;
Weighted calculation submodule, based on being weighted to object to be assessed there are the risk score of the index feature of risk
It calculates, obtains the risk class score of object to be assessed;
Grade determines submodule, for the corresponding relationship according to preset risk class fraction range and risk class, really
Risk class fraction range where the risk class score of fixed object to be assessed, and the risk class fraction range is corresponding
Risk class is determined as the risk evaluation result of object to be assessed.
Further, weighted calculation submodule is also used to:
The risk class score of object to be assessed is calculated according to following formula:
Wherein, P is the risk class score of object to be assessed, kiFor i-th, there are the default power of the index feature of risk
Weight, xiFor i-th there are the risk score of the index feature of risk, n is the quantity of the index feature there are risk, and β is default
Adjustment parameter.
Specific restriction about the risk assessment device based on micro- expression may refer to above for based on micro- expression
The restriction of methods of risk assessment, details are not described herein.Modules in the above-mentioned risk assessment device based on micro- expression can be complete
Portion or part are realized by software, hardware and combinations thereof.Above-mentioned each module can be embedded in the form of hardware or independently of calculating
In processor in machine equipment, it can also be stored in a software form in the memory in computer equipment, in order to processor
It calls and executes the corresponding operation of the above modules.
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 database of machine equipment is used to store the first video data and the second video data of object to be assessed.The net of the computer equipment
Network interface is used to communicate with external terminal by network connection.To realize a kind of base when the computer program is executed by processor
In the methods of risk assessment of micro- expression.
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 are realized in above-described embodiment when executing computer program based on micro- expression
Methods of risk assessment the step of, such as step S1 shown in Fig. 2 to step S6.Alternatively, when processor executes computer program
Realize the function of each module/unit of the risk assessment device based on micro- expression in above-described embodiment, such as module 61 shown in Fig. 6
To the function of module 66.To avoid repeating, 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
The methods of risk assessment based on micro- expression in above method embodiment is realized when program is executed by processor, alternatively, the computer
Each module/unit in the risk assessment device based on micro- expression is realized in above-mentioned apparatus embodiment when program is executed by processor
Function.To avoid repeating, details are not described herein again.
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), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct 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 methods of risk assessment based on micro- expression, which is characterized in that the methods of risk assessment includes:
Obtain the first video data of object to be assessed, wherein first video data is that the object to be assessed is answered in advance
If underlying issue video data;
Micro- Expression Recognition is carried out to first video data using preset micro- Expression Recognition model, it is described to be assessed right to obtain
The micro- expression data of the first of elephant, wherein preset micro- Expression Recognition model identifies default from first video data
Index feature and every kind of index feature characteristic, first micro- expression data includes every kind of index
The characteristic of feature;
According to the characteristic of every kind of index feature in described first micro- expression data, the face of the object to be assessed is established
The micro- expression base-line data in portion, wherein the micro- expression base-line data of face includes the regime values model of every kind of index feature
It encloses;
Obtain the second video data of the object to be assessed, wherein second video data returns for the object to be assessed
Answer the video data of preset evaluation problem;
Micro- Expression Recognition is carried out to second video data using micro- Expression Recognition model, obtains the object to be assessed
Second micro- expression data, wherein preset micro- Expression Recognition model identifies the finger from second video data
The characteristic to be identified of feature and every kind of index feature is marked, second micro- expression data includes every kind of finger
Mark the characteristic to be identified of feature;
Using the regime values range of every kind of index feature in the micro- expression base-line data of the face, to described second micro- table
The characteristic to be identified of every kind of index feature carries out risk assessment in feelings data, obtains the risk of the object to be assessed
Assessment result.
2. methods of risk assessment as described in claim 1, which is characterized in that preset micro- Expression Recognition model includes people
Face detection model, mood discrimination model, head pose identification model, blink detection model and iris edge detection model, it is described
Preset index feature includes muscle movement feature, head pose feature, blink feature and eye movement variation characteristic, and the use is pre-
If micro- Expression Recognition model micro- Expression Recognition is carried out to first video data, obtain the object to be assessed first is micro-
Expression data includes:
According to preset extracting mode, the video frame images of preset quantity are extracted from first video data;
Using the Face datection model, Face datection is carried out to the video frame images, is extracted in the video frame images
Face picture;
The face picture is inputted into the mood discrimination model, micro- Expression Recognition is carried out, obtains the every of the object to be assessed
The characteristic of the kind muscle movement feature;
The face picture is inputted into the head pose identification model, carries out head pose identification, it is described to be assessed right to obtain
The characteristic of the head pose feature of every kind of elephant;
The face picture is inputted into the blink detection model, carries out number of winks detection, obtains frequency of wink, and will be described
Characteristic of the frequency of wink as the blink feature of the object to be assessed;
The face picture is inputted into the iris edge detection model, carries out eye movement variation detection, it is described to be assessed right to obtain
The characteristic of the eye movement variation characteristic of every kind of elephant;
By the characteristic of every kind of muscle movement feature, characteristic, the blink of every kind of head pose feature
The characteristic of feature and the characteristic of every kind of eye variation characteristic are as described first micro- expression data.
3. methods of risk assessment as described in claim 1, which is characterized in that described according to every in described first micro- expression data
The characteristic of the kind index feature, the micro- expression base-line data of face for establishing the object to be assessed include:
Obtain every kind of corresponding default minimum scale coefficient of index feature and default maximum in described first micro- expression data
Proportionality coefficient;
For index feature described in every kind, calculate the index feature characteristic it is corresponding with the index feature it is described it is default most
The characteristic of the first product and the index feature between small scale coefficient it is corresponding with the index feature it is described it is default most
The second product between large scale coefficient;
By the value range between the second product described in corresponding first sum of products of every kind of index feature, it is determined as every
The regime values range of the kind index feature obtains the micro- expression base-line data of face of the object to be assessed.
4. methods of risk assessment as described in any one of claims 1 to 3, which is characterized in that described to use the micro- table of face
The regime values range of every kind of index feature in feelings base-line data, to every kind of index in described second micro- expression data
The characteristic to be identified of feature carries out risk assessment, and the risk evaluation result for obtaining the object to be assessed includes:
For index feature described in every kind, if the characteristic to be identified of the index feature meets the regime values of the index feature
The requirement of range then confirms the index feature devoid of risk of the object to be assessed, otherwise, confirms being somebody's turn to do for the object to be assessed
There are risks for index feature;
For index feature described in every kind, if there are risks for the index feature, the characteristic to be identified of the index feature is calculated
According to the ratio of the regime values range beyond the index feature, and the index of the object to be assessed according to the ratio-dependent is special
The risk score of sign;
To the object to be assessed, there are the risk scores of the index feature of risk to be weighted, and obtains described to be evaluated
Estimate the risk class score of object;
According to the corresponding relationship of preset risk class fraction range and risk class, the risk etc. of the object to be assessed is determined
Risk class fraction range where grade score, and the corresponding risk class of risk class fraction range is determined as the wind
Dangerous assessment result.
5. methods of risk assessment as claimed in claim 4, which is characterized in that described to the object to be assessed, there are risks
The risk score of the index feature is weighted, and the risk class score for obtaining the object to be assessed includes:
The risk class score of the object to be assessed is calculated according to following formula:
Wherein, P is the risk class score, kiFor i-th, there are the default weight of the index feature of risk, xiIt is i-th
A there are the risk score of the index feature of risk, n is the quantity of the index feature there are risk, and β is preset
Adjustment parameter.
6. a kind of risk assessment device based on micro- expression, which is characterized in that the risk assessment device includes:
First obtains module, for obtaining the first video data of object to be assessed, wherein first video data is described
Object to be assessed answers the video data of preset underlying issue;
First identification module, for carrying out micro- expression knowledge to first video data using preset micro- Expression Recognition model
Not, first micro- expression data of the object to be assessed is obtained, wherein preset micro- Expression Recognition model is from described first
The characteristic of preset index feature and every kind of index feature, first micro- expression number are identified in video data
According to the characteristic for including every kind of index feature;
Baseline establishes module, for the characteristic according to every kind of index feature in described first micro- expression data, establishes
The micro- expression base-line data of face of the object to be assessed, wherein the micro- expression base-line data of face includes every kind of finger
Mark the regime values range of feature;
Second obtains module, for obtaining the second video data of the object to be assessed, wherein second video data is
The object to be assessed answers the video data of preset evaluation problem;
Second identification module, for carrying out micro- Expression Recognition to second video data using micro- Expression Recognition model,
Obtain second micro- expression data of the object to be assessed, wherein preset micro- Expression Recognition model is regarded from described second
Frequency identifies the characteristic to be identified of the index feature and every kind of index feature, second micro- expression in
Data include the characteristic to be identified of every kind of index feature;
Risk evaluation module, for using the regime values model of every kind of index feature in the micro- expression base-line data of the face
It encloses, risk assessment is carried out to the characteristic to be identified of every kind of index feature in described second micro- expression data, obtains institute
State the risk evaluation result of object to be assessed.
7. risk assessment device as claimed in claim 6, which is characterized in that preset micro- Expression Recognition model includes people
Face detection model, mood discrimination model, head pose identification model, blink detection model and iris edge detection model, it is described
Preset index feature includes muscle movement feature, head pose feature, blink feature and eye movement variation characteristic, and described first knows
Other module includes:
Frame image zooming-out submodule, for extracting preset quantity from first video data according to preset extracting mode
Video frame images;
Face datection submodule carries out Face datection to the video frame images, extracts for using the Face datection model
Face picture in the video frame images;
Mood differentiates submodule, for the face picture to be inputted the mood discrimination model, carries out micro- Expression Recognition, obtains
The characteristic of the muscle movement feature of every kind of the object to be assessed;
Head pose identifies submodule, for the face picture to be inputted the head pose identification model, carries out head appearance
State identification, obtains the characteristic of every kind of head pose feature of the object to be assessed;
Blink detection submodule carries out number of winks detection, obtains for the face picture to be inputted the blink detection model
To frequency of wink, and using the frequency of wink as the characteristic of the blink feature of the object to be assessed;
Eye movement changes detection sub-module, for the face picture to be inputted the iris edge detection model, carries out eye movement change
Change detection, obtains the characteristic of every kind of eye movement variation characteristic of the object to be assessed;
Micro- expression data generates submodule, for by the characteristic of every kind of muscle movement feature, every kind of head appearance
The characteristic of the characteristic of state feature, the characteristic of the blink feature and every kind of eye variation characteristic is as institute
State first micro- expression data.
8. risk assessment device as claimed in claim 6, which is characterized in that the baseline establishes module and includes:
Coefficient acquisition submodule, for obtaining in described first micro- expression data the corresponding default minimum of every kind of index feature
Proportionality coefficient and default maximum ratio coefficient;
Data expansion submodule calculates the characteristic and the index of the index feature for being directed to every kind of index feature
The characteristic and the index of the first product and the index feature between the corresponding default minimum scale coefficient of feature
The second product between the corresponding default maximum ratio coefficient of feature;
Baseline determines submodule, for will be between the second product described in corresponding first sum of products of every kind of index feature
Value range, be determined as the regime values range of every kind of index feature, obtain the face of the object to be assessed
Micro- expression base-line data.
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
The step of methods of risk assessment based on micro- expression described in 5 any one.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In risk of the realization based on micro- expression as described in any one of claim 1 to 5 is commented when the computer program is executed by processor
The step of estimating method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811182176.8A CN109472206B (en) | 2018-10-11 | 2018-10-11 | Risk assessment method, device, equipment and medium based on micro-expressions |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811182176.8A CN109472206B (en) | 2018-10-11 | 2018-10-11 | Risk assessment method, device, equipment and medium based on micro-expressions |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109472206A true CN109472206A (en) | 2019-03-15 |
CN109472206B CN109472206B (en) | 2023-07-07 |
Family
ID=65664675
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811182176.8A Active CN109472206B (en) | 2018-10-11 | 2018-10-11 | Risk assessment method, device, equipment and medium based on micro-expressions |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109472206B (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110457432A (en) * | 2019-07-04 | 2019-11-15 | 平安科技(深圳)有限公司 | Interview methods of marking, device, equipment and storage medium |
CN110490424A (en) * | 2019-07-23 | 2019-11-22 | 阿里巴巴集团控股有限公司 | A kind of method and apparatus of the progress risk assessment based on convolutional neural networks |
CN111192150A (en) * | 2019-12-23 | 2020-05-22 | 中国平安财产保险股份有限公司 | Method, device and equipment for processing vehicle insurance agent business and storage medium |
CN111223078A (en) * | 2019-12-31 | 2020-06-02 | 河南裕展精密科技有限公司 | Method for determining defect grade and storage medium |
CN111339859A (en) * | 2020-02-17 | 2020-06-26 | 出门问问信息科技有限公司 | Multi-modal risk control method and device and computer-readable storage medium |
CN111540440A (en) * | 2020-04-23 | 2020-08-14 | 深圳市镜象科技有限公司 | Psychological examination method, device, equipment and medium based on artificial intelligence |
CN111553311A (en) * | 2020-05-13 | 2020-08-18 | 吉林工程技术师范学院 | Micro-expression recognition robot and control method thereof |
CN111708939A (en) * | 2020-05-29 | 2020-09-25 | 平安科技(深圳)有限公司 | Push method and device based on emotion recognition, computer equipment and storage medium |
CN112200462A (en) * | 2020-10-13 | 2021-01-08 | 中国银行股份有限公司 | Risk assessment method and device |
CN112633239A (en) * | 2020-12-31 | 2021-04-09 | 中国工商银行股份有限公司 | Micro-expression identification method and device |
CN112818754A (en) * | 2021-01-11 | 2021-05-18 | 广州番禺职业技术学院 | Learning concentration degree judgment method and device based on micro-expressions |
CN112819609A (en) * | 2021-02-24 | 2021-05-18 | 深圳前海微众银行股份有限公司 | Risk assessment method, apparatus, computer-readable storage medium, and program product |
CN113743335A (en) * | 2021-09-08 | 2021-12-03 | 平安科技(深圳)有限公司 | Method, device, computer and medium for risk identification of gaze data |
WO2022119536A1 (en) * | 2020-12-01 | 2022-06-09 | Turkiye Garanti Bankasi Anonim Sirketi | A system for detecting persons who may perform suspicious actions |
CN115050081A (en) * | 2022-08-12 | 2022-09-13 | 平安银行股份有限公司 | Expression sample generation method, expression recognition method and device and terminal equipment |
CN115798022A (en) * | 2023-02-07 | 2023-03-14 | 南京思优普信息科技有限公司 | Artificial intelligence identification method based on feature extraction |
CN112200462B (en) * | 2020-10-13 | 2024-04-26 | 中国银行股份有限公司 | Risk assessment method and risk assessment device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2016149063A (en) * | 2015-02-13 | 2016-08-18 | オムロン株式会社 | Emotion estimation system and emotion estimation method |
CN107480622A (en) * | 2017-08-07 | 2017-12-15 | 深圳市科迈爱康科技有限公司 | Micro- expression recognition method, device and storage medium |
CN108537160A (en) * | 2018-03-30 | 2018-09-14 | 平安科技(深圳)有限公司 | Risk Identification Method, device, equipment based on micro- expression and medium |
-
2018
- 2018-10-11 CN CN201811182176.8A patent/CN109472206B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2016149063A (en) * | 2015-02-13 | 2016-08-18 | オムロン株式会社 | Emotion estimation system and emotion estimation method |
CN107480622A (en) * | 2017-08-07 | 2017-12-15 | 深圳市科迈爱康科技有限公司 | Micro- expression recognition method, device and storage medium |
CN108537160A (en) * | 2018-03-30 | 2018-09-14 | 平安科技(深圳)有限公司 | Risk Identification Method, device, equipment based on micro- expression and medium |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110457432B (en) * | 2019-07-04 | 2023-05-30 | 平安科技(深圳)有限公司 | Interview scoring method, interview scoring device, interview scoring equipment and interview scoring storage medium |
CN110457432A (en) * | 2019-07-04 | 2019-11-15 | 平安科技(深圳)有限公司 | Interview methods of marking, device, equipment and storage medium |
CN110490424A (en) * | 2019-07-23 | 2019-11-22 | 阿里巴巴集团控股有限公司 | A kind of method and apparatus of the progress risk assessment based on convolutional neural networks |
CN111192150A (en) * | 2019-12-23 | 2020-05-22 | 中国平安财产保险股份有限公司 | Method, device and equipment for processing vehicle insurance agent business and storage medium |
CN111192150B (en) * | 2019-12-23 | 2023-07-25 | 中国平安财产保险股份有限公司 | Method, device, equipment and storage medium for processing vehicle danger-giving agent service |
CN111223078A (en) * | 2019-12-31 | 2020-06-02 | 河南裕展精密科技有限公司 | Method for determining defect grade and storage medium |
CN111223078B (en) * | 2019-12-31 | 2023-09-26 | 富联裕展科技(河南)有限公司 | Method for determining flaw level and storage medium |
CN111339859A (en) * | 2020-02-17 | 2020-06-26 | 出门问问信息科技有限公司 | Multi-modal risk control method and device and computer-readable storage medium |
CN111540440A (en) * | 2020-04-23 | 2020-08-14 | 深圳市镜象科技有限公司 | Psychological examination method, device, equipment and medium based on artificial intelligence |
CN111540440B (en) * | 2020-04-23 | 2021-01-15 | 深圳市镜象科技有限公司 | Psychological examination method, device, equipment and medium based on artificial intelligence |
CN111553311A (en) * | 2020-05-13 | 2020-08-18 | 吉林工程技术师范学院 | Micro-expression recognition robot and control method thereof |
CN111708939B (en) * | 2020-05-29 | 2024-04-16 | 平安科技(深圳)有限公司 | Emotion recognition-based pushing method and device, computer equipment and storage medium |
CN111708939A (en) * | 2020-05-29 | 2020-09-25 | 平安科技(深圳)有限公司 | Push method and device based on emotion recognition, computer equipment and storage medium |
CN112200462A (en) * | 2020-10-13 | 2021-01-08 | 中国银行股份有限公司 | Risk assessment method and device |
CN112200462B (en) * | 2020-10-13 | 2024-04-26 | 中国银行股份有限公司 | Risk assessment method and risk assessment device |
WO2022119536A1 (en) * | 2020-12-01 | 2022-06-09 | Turkiye Garanti Bankasi Anonim Sirketi | A system for detecting persons who may perform suspicious actions |
CN112633239A (en) * | 2020-12-31 | 2021-04-09 | 中国工商银行股份有限公司 | Micro-expression identification method and device |
CN112818754A (en) * | 2021-01-11 | 2021-05-18 | 广州番禺职业技术学院 | Learning concentration degree judgment method and device based on micro-expressions |
CN112819609A (en) * | 2021-02-24 | 2021-05-18 | 深圳前海微众银行股份有限公司 | Risk assessment method, apparatus, computer-readable storage medium, and program product |
CN113743335B (en) * | 2021-09-08 | 2024-03-22 | 平安科技(深圳)有限公司 | Method, device, computer and medium for risk identification of gaze data |
CN113743335A (en) * | 2021-09-08 | 2021-12-03 | 平安科技(深圳)有限公司 | Method, device, computer and medium for risk identification of gaze data |
CN115050081B (en) * | 2022-08-12 | 2022-11-25 | 平安银行股份有限公司 | Expression sample generation method, expression recognition method and device and terminal equipment |
CN115050081A (en) * | 2022-08-12 | 2022-09-13 | 平安银行股份有限公司 | Expression sample generation method, expression recognition method and device and terminal equipment |
CN115798022A (en) * | 2023-02-07 | 2023-03-14 | 南京思优普信息科技有限公司 | Artificial intelligence identification method based on feature extraction |
Also Published As
Publication number | Publication date |
---|---|
CN109472206B (en) | 2023-07-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109472206A (en) | Methods of risk assessment, device, equipment and medium based on micro- expression | |
EP3885965B1 (en) | Image recognition method based on micro facial expressions, apparatus and related device | |
CN109146856A (en) | Picture quality assessment method, device, computer equipment and storage medium | |
CN108829900A (en) | A kind of Research on face image retrieval based on deep learning, device and terminal | |
CN108537160A (en) | Risk Identification Method, device, equipment based on micro- expression and medium | |
CN109614488B (en) | Text classification and image recognition-based distribution network live working condition judgment method | |
CN107798318A (en) | The method and its device of a kind of happy micro- expression of robot identification face | |
CN107145857A (en) | Face character recognition methods, device and method for establishing model | |
CN105374055A (en) | Image processing method and device | |
CN106295591A (en) | Gender identification method based on facial image and device | |
CN104915658B (en) | A kind of emotion component analyzing method and its system based on emotion Distributed learning | |
CN111414858B (en) | Face recognition method, target image determining device and electronic system | |
Paul et al. | Extraction of facial feature points using cumulative histogram | |
CN113139439B (en) | Online learning concentration evaluation method and device based on face recognition | |
CN112883867A (en) | Student online learning evaluation method and system based on image emotion analysis | |
CN106650555A (en) | Real person verifying method and system based on machine learning | |
KR20210090456A (en) | Image-based Posture Preservation Virtual Fitting System Supporting Multi-Poses | |
CN112580572A (en) | Training method of multi-task recognition model, using method, equipment and storage medium | |
CN107392134A (en) | A kind of face identification method and system based on joint piecemeal | |
CN112818796B (en) | Intelligent gesture distinguishing method and storage device suitable for online prison scene | |
Jearanaitanakij et al. | Predicting short trend of stocks by using convolutional neural network and candlestick patterns | |
Lv et al. | A spontaneous facial expression recognition method using head motion and AAM features | |
CN110399818A (en) | A kind of method and apparatus of risk profile | |
CN115937928A (en) | Learning state monitoring method and system based on multi-vision feature fusion | |
Harit et al. | Facial emotions recognition using gabor transform and facial animation parameters with neural networks |
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 |