CN109858405A - Satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression - Google Patents
Satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression Download PDFInfo
- Publication number
- CN109858405A CN109858405A CN201910043295.3A CN201910043295A CN109858405A CN 109858405 A CN109858405 A CN 109858405A CN 201910043295 A CN201910043295 A CN 201910043295A CN 109858405 A CN109858405 A CN 109858405A
- Authority
- CN
- China
- Prior art keywords
- images
- recognized
- segment
- negative feeling
- target
- 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
- 230000014509 gene expression Effects 0.000 title claims abstract description 141
- 238000011156 evaluation Methods 0.000 title claims abstract description 61
- 230000036651 mood Effects 0.000 claims abstract description 128
- 239000012634 fragment Substances 0.000 claims description 23
- 238000004590 computer program Methods 0.000 claims description 17
- 235000021167 banquet Nutrition 0.000 claims description 11
- 238000000034 method Methods 0.000 abstract description 25
- 230000001815 facial effect Effects 0.000 description 14
- 230000006870 function Effects 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 8
- 210000004709 eyebrow Anatomy 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 7
- 230000007935 neutral effect Effects 0.000 description 7
- 238000012549 training Methods 0.000 description 5
- 230000008451 emotion Effects 0.000 description 4
- 238000010195 expression analysis Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008921 facial expression Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000002996 emotional effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 210000000744 eyelid Anatomy 0.000 description 1
- 210000004209 hair Anatomy 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Abstract
The present invention discloses a kind of satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression.This method comprises: obtaining the video data of target customer, images to be recognized each in video data is identified using micro- Expression Recognition model, obtains mood attribute；The first frame images to be recognized that mood attribute is negative feeling is determined as benchmark image；Target video segment is formed based on the images to be recognized in preset time period after benchmark image；Based on the corresponding mood attribute of images to be recognized each in target video segment, negative feeling probability is obtained；If negative feeling probability is greater than predetermined probabilities threshold value, target video segment is determined as negative feeling segment；Negative feeling segment is handled using dissatisfaction formula, obtains target not full value；According to target, full value does not inquire grading system tables of data, obtains target grading system.The target grading system that this method obtains has more objectivity, can the more intuitive service satisfaction for embodying target customer.
Description
Technical field
The present invention relates to micro- Expression Recognition technical field more particularly to a kind of satisfaction evaluation method based on micro- expression,
Device, equipment and storage medium.
Background technique
In the financial institutions such as current bank, security and insurance or other service organizations, it will usually configure satisfaction evaluation
System carries out satisfaction marking to acquire client to service provided by financial institution or other service organizations, to be based on
The satisfaction marking result of client improves service, to improve service quality.Current satisfaction evaluation system include scoring modules and
Server corresponding with scoring modules is provided with satisfied and dissatisfied equal keys or very satisfied, full on the scoring modules
Meaning, general and dissatisfied etc. keys.Client can be by clicking corresponding key on scoring modules, and realization satisfaction is given a mark, so that
Server gets corresponding satisfaction marking result.In current satisfaction evaluation system, client is set in advance by scoring modules
The key set realizes satisfaction marking, does not have specific limit to the corresponding satisfaction of different key, makes its satisfaction marking knot
The objectivity of fruit is poor, can not be truly reflected client to the satisfaction of service.
Summary of the invention
The embodiment of the present invention provides a kind of satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression, with
It is poor to solve the problems, such as to realize in satisfaction scoring process that there are objectivity by the keys of scoring modules.
A kind of satisfaction evaluation method based on micro- expression, comprising:
Video data when target customer receives service is obtained, the video data includes an at least frame images to be recognized,
Each corresponding timestamp of the images to be recognized；
Each images to be recognized is identified using micro- Expression Recognition model, it is corresponding to obtain the images to be recognized
Mood attribute；
It is according to the sequence of the timestamp, images to be recognized described in first frame of the mood attribute for negative feeling is true
It is set to benchmark image；
Images to be recognized based on the timestamp after the timestamp of the benchmark image in preset time period is formed
Target video segment；
Based on the corresponding mood attribute of the images to be recognized each in the target video segment, the target view is obtained
The corresponding negative feeling probability of frequency segment；
If the corresponding negative feeling probability of the target video segment is greater than predetermined probabilities threshold value, by the target video
Segment is determined as negative feeling segment, is based on the negative feeling fragment update benchmark image, repeats described based on described
Images to be recognized of the timestamp after the timestamp of the benchmark image in preset time period forms target video segment；
The negative feeling segment is handled using dissatisfaction formula, obtains target not full value；
According to the target, full value does not inquire grading system tables of data, obtains target grading system.
A kind of satisfaction evaluation device based on micro- expression, comprising:
Video data obtains module, for obtaining video data when target customer receives service, the video data packet
Include an at least frame images to be recognized, each corresponding timestamp of the images to be recognized；
Mood attribute obtains module, for being identified using micro- Expression Recognition model to each images to be recognized,
Obtain the corresponding mood attribute of the images to be recognized；
The mood attribute is negative feeling for the sequence according to the timestamp by benchmark image determining module
Images to be recognized described in first frame is determined as benchmark image；
Target video segment forms module, for being preset after the timestamp of the benchmark image based on the timestamp
Images to be recognized in period forms target video segment；
Negative feeling probability obtains module, for corresponding based on the images to be recognized each in the target video segment
Mood attribute, obtain the corresponding negative feeling probability of the target video segment；
Negative feeling segment determining module, if being greater than for the corresponding negative feeling probability of the target video segment default
The target video segment is then determined as negative feeling segment by probability threshold value, is based on the negative feeling fragment update benchmark
Image repeats described to be identified in preset time period after the timestamp of the benchmark image based on the timestamp
Image forms target video segment；
Full value does not obtain module to target, for being handled using dissatisfaction formula the negative feeling segment, obtains
Take target not full value；
Target grading system obtains module, for according to the target, full value not to inquire grading system tables of data, acquisition mesh
Mark grading system.
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 above-mentioned satisfaction based on micro- expression when executing the computer program
Evaluation method.
A kind of computer readable storage medium, the computer-readable recording medium storage have computer program, the meter
Calculation machine program realizes the above-mentioned satisfaction evaluation method based on micro- expression when being executed by processor.
In the above-mentioned satisfaction evaluation method, apparatus based on micro- expression, equipment and storage medium, by being connect to target customer
Images to be recognized in video data when being serviced carries out micro- Expression analysis, quickly determines the corresponding feelings of each images to be recognized
Thread attribute, to analyze the emotional change situation of target customer in real time.It is the images to be recognized shape of negative attributes based on mood attribute
Corresponding target video segment is obtained at benchmark image, and based on benchmark image, target video segment is determined as satisfaction
The basic unit of analysis facilitates the data volume for reducing analysis, reduces interference, improves analysis efficiency and accuracy.According to target
The mood attribute of each images to be recognized in video clip determines its negative feeling probability, and is greater than in advance in negative feeling probability
If when probability threshold value, target video segment is determined as negative feeling segment, so that negative feeling segment can more objectively reflect
The dissatisfaction of target customer.Negative feeling segment is handled using dissatisfaction formula, can quick obtaining it is corresponding
Target not full value, full value objectivity is not stronger for the target, can intuitively reflect the dissatisfaction of target customer.It is discontented according to target
Value inquiry grading system tables of data, can quick obtaining target grading system so that target grading system have more objectivity, can be more
The intuitive service satisfaction for embodying target customer to the service received.
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 satisfaction evaluation method in one embodiment of the invention based on micro- expression；
Fig. 2 is a flow chart of the satisfaction evaluation method in one embodiment of the invention based on micro- expression；
Fig. 3 is another flow chart of the satisfaction evaluation method in one embodiment of the invention based on micro- expression；
Fig. 4 is another flow chart of the satisfaction evaluation method in one embodiment of the invention based on micro- expression；
Fig. 5 is another flow chart of the satisfaction evaluation method in one embodiment of the invention based on micro- expression；
Fig. 6 is another flow chart of the satisfaction evaluation method in one embodiment of the invention based on micro- expression；
Fig. 7 is a schematic diagram of the satisfaction evaluation square law device in one embodiment of the invention based on micro- expression；
Fig. 8 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.
Satisfaction evaluation method provided in an embodiment of the present invention based on micro- expression, should the satisfaction evaluation based on micro- expression
Method can be using in application environment as shown in Figure 1.Specifically, the satisfaction evaluation method based on micro- expression is somebody's turn to do to apply in client
In service evaluation system, which includes server as shown in Figure 1, monitor terminal and terminal of attending a banquet, prison
Control terminal and terminal of attending a banquet are communicated with server by network, are being received for acquiring target customer by monitor terminal
Video data in service process carries out micro- Expression Recognition to video data, determines target visitor according to the mood attribute identified
The dissatisfaction at family, to determine target grading system according to dissatisfaction, to obtain objective target grading system.Wherein,
Terminal of attending a banquet is user terminal, refers to corresponding with server, provides the program of local service for client, it is mountable but it is unlimited
In on various personal computers, laptop, smart phone, tablet computer and portable wearable device.Server can be with
It is realized with the server cluster of the either multiple server compositions of independent server.In the present embodiment, monitor terminal is to use
In the terminal of the video data of acquisition target customer, i.e. terminal used in target customer, be specifically as follows picture pick-up device or
Other carry the terminal of camera.Terminal of attending a banquet is the financial institution of service goal client or the people that attends a banquet of other service organizations
The terminal that member or other business personnels use.
In one embodiment, it as shown in Fig. 2, providing a kind of satisfaction evaluation method based on micro- expression, answers in this way
It is illustrated, includes the following steps: for the server in Fig. 1
S201: obtaining video data when target customer receives service, and video data includes an at least frame images to be recognized,
The corresponding timestamp of each images to be recognized.
Specifically, receive seat personnel or other business people of financial institution or other service organizations in target customer
When the service that member provides, the video data of target customer is acquired in real time using monitor terminal, and the video data is sent to clothes
Business device, so that server carries out micro- Expression analysis to the video data.Wherein, target customer refers to the visitor for receiving service
Family this time needs to carry out the client of satisfaction evaluation.Service provided by seat personnel or other business personnels include but
It is not limited to basic business and handles service (such as user account number is opened an account) and client's return visit service.
The video data includes an at least frame images to be recognized, which is the single frames figure to form video data
Picture, i.e., the single width image frame of minimum unit in video data.The corresponding recording time of each images to be recognized, the recording time
To take the corresponding timestamp of each images to be recognized.In the present embodiment, monitor terminal is when shooting video data, according to pre-
The filming frequency being first arranged is shot.The filming frequency refers to the quantity of the single-frame images shot in the unit time, for example, if
It is 5-10 frame/second that its filming frequency f, which is arranged, then monitor terminal shooting 5-10 frame images to be recognized per second.
S202: identifying each images to be recognized using micro- Expression Recognition model, and it is corresponding to obtain images to be recognized
Mood attribute.
Wherein, micro- Expression Recognition model is the model of the micro- expression of face in images to be recognized for identification.In the present embodiment,
Micro- Expression Recognition model be by capture images to be recognized in user face local feature, and according to local feature determine to
The each target face motor unit for identifying face in image, determines its feelings further according to the target face motor unit identified
The model of thread.The corresponding mood attribute of images to be recognized refers to according to the micro- expression type identified in images to be recognized, determines
Mood attribute corresponding with the micro- expression type.Micro- expression type be using micro- Expression Recognition model to images to be recognized into
The type of the micro- expression of face determined after row identification, identifies images to be recognized specifically by micro- Expression Recognition model,
The micro- expression determined according to the target face motor unit identified.Specifically, server first uses micro- Expression Recognition model pair
Each images to be recognized carries out micro- Expression Recognition, then, true according to micro- expression type to determine its corresponding micro- expression type
Its fixed corresponding mood attribute.The mood attribute includes active mood, negative feeling and neutral mood.
In the present embodiment, micro- Expression Recognition model can be the neural network recognization model based on deep learning, can also be with
It is the local identification model based on classification, can also be based on local binary patterns (Local Binary Pattern, LBP)
Local Emotion identification model.In the present embodiment, micro- Expression Recognition model is the local identification model based on classification, micro- Expression Recognition
It include each face in training image data by collecting a large amount of training image data in advance when model is trained in advance
The positive sample of motor unit and the negative sample of Facial action unit are trained training image data by sorting algorithm, obtain
Take micro- Expression Recognition model.In the present embodiment, it can be and a large amount of training image data are instructed by svm classifier algorithm
Practice, to get SVM classifier corresponding with multiple Facial action units.For example, it may be 39 Facial action units are corresponding
39 SVM classifiers, be also possible to corresponding 54 SVM classifiers of 54 Facial action units, be trained training figure
As the positive sample and negative sample of the different Facial action units that include in data are more, then the SVM classifier quantity got is got over
It is more.It is to be appreciated that the SVM classifier got is got over by multiple SVM classifiers to be formed in micro- Expression Recognition model
More, then micro- expression type that the micro- Expression Recognition model formed identifies is more accurate.It is corresponding with 54 Facial action units
SVM classifier is formed by for micro- Expression Recognition model, may recognize that 54 kinds of micro- expressions using this micro- Expression Recognition model
Type, such as may recognize that include to like, is interested, pleasantly surprised, expecting ... aggressive, conflict, insult, suspection and fear etc. 54 kinds
Micro- expression type.
S203: according to the sequence of timestamp, the first frame images to be recognized that mood attribute is negative feeling is determined as base
Quasi- image.
Specifically, mood attribute in video data is by sequence of the server according to the timestamp of each images to be recognized
The first frame images to be recognized of negative feeling is determined as benchmark image, to carry out subsequent mood analysis based on the benchmark image.
It is to be appreciated that after carrying out micro- Expression analysis to each frame images to be recognized in video data, according to the suitable of timestamp
The first frame images to be recognized that mood attribute is negative feeling is determined as benchmark image, and is started based on the benchmark image by sequence
One monitoring thread, the corresponding mood of images to be recognized in the preset time period after timestamp for monitoring the benchmark image
Attribute.For example, the sequence according to timestamp, first carries out Emotion identification to the 1st frame images to be recognized, if its mood attribute is passiveness
Mood is then determined as benchmark image；If not being negative feeling, then it is straight Emotion identification ... successively to be carried out to the 2nd frame images to be recognized
To the images to be recognized that first mood attribute is negative feeling is recognized, which is determined as benchmark image.
S204: target is formed based on images to be recognized of the timestamp after the timestamp of benchmark image in preset time period
Video clip.
Wherein, preset time period is the pre-set period, for example, the settable preset time period is in the present embodiment
10 seconds.Specifically, server is carrying out video data to analyze and determine that a certain frame images to be recognized is base after benchmark image
In the benchmark image start a monitoring thread, for acquisition time stamp in the preset time period after the benchmark image wait know
Other image, so as to the target video segment formed based on all images to be recognized in the benchmark image and preset time period.It should
Target video segment is the segment formed based on the images to be recognized in a benchmark image and its later preset time period.This implementation
In example, using the target video segment as the basic unit of target customer's Analysis of Satisfaction in receiving service process, to exclude
Mood attribute is the interference of the images to be recognized of neutral mood or active mood before benchmark image.For example, a certain reference map
The timestamp of picture be 11 points 01 second 10 minutes, preset time period is 20 seconds, then is specifically managed after the benchmark image in preset time period
Solution for 11 points 01 second 10 minutes -11 point 10 minutes and 21 seconds, it is all between 01 second 10 minutes -11 point 10 minutes and 21 seconds at 11 points based on timestamp
Images to be recognized forms target video segment.
S205: based on the corresponding mood attribute of images to be recognized each in target video segment, target video segment is obtained
Corresponding negative feeling probability.
Wherein, negative feeling probability refers to that mood attribute in target video segment is the general of the images to be recognized of negative feeling
Rate.Specifically, server determines that mood attribute is based on the corresponding mood attribute of images to be recognized each in target video segment
The ratio-dependent is that the corresponding negative feeling of target video segment is general by ratio shared by the images to be recognized of negative feeling
Rate.It is to be appreciated that the corresponding negative feeling probability of the target video segment can reflect it is pre- from the timestamp of benchmark image
If in the period, target customer is in the shape probability of state of negative feeling in receiving service process, it can more objectively reflect mesh
Mark the dissatisfaction of client.For example, the negative feeling probability is 30%, then illustrate pre- the timestamp from benchmark image
If in the period, target customer has 30% time to be in the state of negative feeling, reflect service of the target customer to being received
Dissatisfaction.
S206: if the corresponding negative feeling probability of target video segment is greater than predetermined probabilities threshold value, by target video piece
Section is determined as negative feeling segment, is based on negative feeling fragment update benchmark image, repeats based on timestamp in reference map
Images to be recognized after the timestamp of picture in preset time period forms target video segment.
Wherein, predetermined probabilities threshold value is pre-set for assessing whether target video segment is passive video clip
Probability threshold value.The predetermined probabilities threshold value can be independently arranged according to the actual situation, for example, the predetermined probabilities threshold value may be configured as
50%.In the present embodiment, if the corresponding negative feeling probability of target video segment is greater than predetermined probabilities threshold value, illustrate in benchmark
In preset time period after the timestamp of image, the facial expression of target customer is in negative feeling (as detested, disliking, instead
To, it is discontented, indignation, ignore, despise, the moods such as false ... and fear) probability it is larger, illustrate target customer to seat personnel
Or the topic point currently talked about of other attendants or service content are more dissatisfied, therefore, can assert the target video piece
Section is passive video clip.It is to be appreciated that passive video clip refers to that mood attribute is that the images to be recognized of negative feeling accounts for
Bigger target video segment can largely reflect the dissatisfaction of target customer, exclude target customer and receiving clothes
The influence (can not form negative feeling segment at this time) of the negative feeling accidentally occurred during business both can help to improve final
The accuracy of target not full value is calculated, and subsequent calculation amount can be reduced.
Further, after determining a passive video clip in video data, server is also needed according to video data
In, the timestamp of each images to be recognized carries out mood analysis to other images to be recognized after the passiveness video clip, with
New benchmark image is determined, to analyze the images to be recognized after the passiveness video clip.In the present embodiment, step
In S206 based on negative feeling fragment update benchmark image, specifically include the following two kinds implementation:
One is mood attribute is that first images to be recognized of negative feeling is updated to by after negative feeling segment
Benchmark image.
For example, a certain video data has 10000 frame images to be recognized, if the 100th frame waits for if predetermined probabilities threshold value is 50%
Identification image is benchmark image, determines that 100-1100 frame images to be recognized is formed by target video piece based on the benchmark image
The negative feeling probability of section is greater than predetermined probabilities threshold value, then the target video segment is passive video clip, illustrates 100-
In 1100 frame images to be recognized, mood attribute is that the accounting of the images to be recognized of negative feeling is more than 50%, at this point, by the passiveness
Mood attribute is that first images to be recognized of negative feeling is updated to new benchmark image (such as the 1105th frame after video clip
Images to be recognized), and step S204 is repeated based on new benchmark image.It is to be appreciated that the reference map that this determination is new
The mode of picture is simple and convenient.
The second is mood attribute is that the last one images to be recognized of negative feeling is updated to by negative feeling segment
Benchmark image.
For example, a certain video data has 10000 frame images to be recognized, if the 100th frame waits for if predetermined probabilities threshold value is 50%
Identification image is benchmark image, determines that 100-1100 frame images to be recognized is formed by target video piece based on the benchmark image
The negative feeling probability of section is greater than predetermined probabilities threshold value, then the target video segment is passive video clip, illustrates 100-
In 1100 frame images to be recognized, mood attribute is that the accounting of the images to be recognized of negative feeling is more than 50%, at this point, by the passiveness
The last one mood attribute of video clip is that the images to be recognized of negative feeling is updated to new benchmark image (such as the 1095th frame
Images to be recognized, the mood attribute of this few frame images to be recognized of 1096-1100 is not negative feeling at this time), and based on new base
Quasi- image repeats step S204.In this mode, target view is formed by according to new benchmark image (i.e. the 1095th frame)
Frequency segment is 1095-1195 frame images to be recognized, to guarantee the continuity between negative feeling segment.
It is to be appreciated that by the last one mood attribute in negative feeling segment be negative feeling images to be recognized more
When being newly benchmark image, if the images to be recognized in preset time period after the timestamp of new benchmark image forms target view
Frequency segment is also negative feeling segment, then, can there are intersection (such as 1095-1100) between the two negative feeling segments
The merging of the two negative feeling segments is updated to new negative feeling segment, to avoid subsequent the problem of duplicating calculating.
S207: being handled negative feeling segment using dissatisfaction formula, obtains target not full value.
Wherein, dissatisfaction formula is for calculating target customer to the formula of the dissatisfaction of the service received.
Full value is not the dissatisfied angle value obtained after being calculated using dissatisfaction formula all negative feeling segments to target.The mesh
Full value can not objectively respond target customer on the whole and receiving the dissatisfaction in service process mark.It is to be appreciated that mesh
Full value is not higher for mark, illustrates target customer to receive the dissatisfaction in service process higher；Conversely, target not get over by full value
It is low, illustrate target customer to receive the dissatisfaction in service process lower.In the present embodiment, using dissatisfaction formula pair
All negative feeling segments are handled, and to determine target not full value, are realized only from negative feeling, exclude active mood and
The interference of neutral mood, making its calculated target, full value does not objectively respond target customer more to the dissatisfied journey for receiving service
Degree；Moreover, dissatisfaction formula is analyzed using negative feeling segment as analytical unit, a few frame moods accidentally occurred can be excluded
Attribute is the interference of the images to be recognized of negative feeling, to guarantee the finally determining target not accuracy of full value and objective
Property.
S208: according to target, full value does not inquire grading system tables of data, obtains target grading system.
Wherein, the grading system table of comparisons is pre-set for determining the dissatisfied corresponding service grading system of angle value
Tables of data.Target grading system is the grade that finally scores of the target customer to the service received.For example, grading system compares
It may be provided with very satisfied, satisfied, general in table and be unsatisfied with these service grading systems, each service grading system is corresponding
One dissatisfied angle value range.It is to be appreciated that dissatisfied angle value range is smaller, corresponding service grading system is more advanced.Often
The corresponding dissatisfied angle value range of one service grading system can be independently arranged by according to the actual situation.
In the present embodiment, server is calculating target not after full value using dissatisfaction formula, not based on the target
Full value inquires grading system tables of data, to determine the target not dissatisfied angle value range belonging to full value, and by belonging to it not
It is satisfied with the corresponding service grading system of angle value range and is determined as target grading system.The determination of the target grading system is by specific
Standards of grading (different dissatisfied angle value range i.e. in grading system tables of data) determine, so that its appraisal result is with more objective
Property, it can more embody service satisfaction of the target customer to the service received.
Further, after step S205, i.e., based on the corresponding feelings of images to be recognized each in target video segment
Thread attribute, after obtaining the corresponding negative feeling probability of target video segment, the satisfaction evaluation method based on micro- expression is also wrapped
It includes:
S209: if the corresponding negative feeling probability of target video segment is not more than predetermined probabilities threshold value, by benchmark image
Later, mood attribute is that the next frame images to be recognized of negative feeling is determined as benchmark image, repeats and is existed based on timestamp
Images to be recognized after the timestamp of benchmark image in preset time period forms target video segment.
Specifically, if the corresponding negative feeling probability of target video segment is not more than predetermined probabilities threshold value, illustrate in base
In preset time period after the timestamp of quasi- image, the facial expression of target customer be in negative feeling (as detested, it is disagreeable,
Oppose, be discontented, indignation, ignore, despise, the moods such as false ... and fear) probability it is smaller, illustrate target customer not to seat
The topic point or service content that seat personnel or other attendants currently talk about show more unsatisfied facial expression, with
Determine that the target video segment is not passive video clip.In the present embodiment, the corresponding target view of any benchmark image is being determined
, need to be by after the timestamp of benchmark image when frequency segment is not negative feeling segment, mood attribute is the next frame of negative feeling
Images to be recognized is determined as new benchmark image, to carry out subsequent analysis based on new benchmark image, i.e. execution step S204
And its later step.For example, a certain video data has 10000 frame images to be recognized if predetermined probabilities threshold value is 50%, if the
100 frame images to be recognized are benchmark image, determine that 100-1100 frame images to be recognized is formed by mesh based on the benchmark image
The negative feeling probability for marking video clip is not more than predetermined probabilities threshold value, then the target video segment is not passive video clip,
Illustrate in 100-1100 frame images to be recognized, mood attribute be the images to be recognized accounting of negative feeling be less than 50% (or
Person says negligible amounts), at this point, directly that the mood attribute after benchmark image is true for the first frame images to be recognized of negative feeling
Be set to new benchmark image (such as the 105th frame images to be recognized), and based on new benchmark image repeat step S204 and its
Later step carries out global analysis to entire video data to realize.
In satisfaction evaluation method based on micro- expression provided by the present embodiment, when by receiving to service to target customer
Video data in images to be recognized carry out micro- Expression analysis, quickly determine the corresponding mood attribute of each images to be recognized,
To analyze the emotional change situation of target customer in real time.Reference map is formed based on the images to be recognized that mood attribute is negative attributes
Picture, and corresponding target video segment is obtained based on benchmark image, target video segment is determined as to the base of Analysis of Satisfaction
This unit facilitates the data volume for reducing analysis, reduces interference, improves analysis efficiency and accuracy.According to target video segment
In each images to be recognized mood attribute, determine its negative feeling probability, and be greater than predetermined probabilities threshold in negative feeling probability
When value, target video segment is determined as negative feeling segment, so that negative feeling segment can more objectively reflect target customer
Dissatisfaction.Negative feeling segment is handled using dissatisfaction formula, can the corresponding target of quick obtaining it is discontented
Value, full value objectivity is not stronger for the target, can intuitively reflect the dissatisfaction of target customer.According to target, full value inquiry is not commented
Grading data table, can quick obtaining target grading system so that target grading system have more objectivity, more intuitive can embody
Service satisfaction of the target customer to the service received.
In one embodiment, as shown in figure 3, step S202, that is, use micro- Expression Recognition model to each images to be recognized
It is identified, obtains the corresponding mood attribute of images to be recognized, specifically comprise the following steps:
S301: identifying each images to be recognized using micro- Expression Recognition model, obtains at least one identification expression
The corresponding instantaneous confidence level of type.
Wherein, it when identification expression type refers to that the micro- Expression Recognition model of use identifies images to be recognized, recognizes
It belongs to the micro- expression type of preconfigured a certain kind.
It specifically, include multiple SVM classifiers in the preparatory trained micro- Expression Recognition model of server, every SVM divides
A kind of class device Facial action unit for identification.In the present embodiment, includes 54 SVM classifiers in micro- Expression Recognition model, build
Vertical Facial action unit number mapping table, each Facial action unit are indicated with a prespecified number.For example, AU1 is
Interior eyebrow raises up, AU2 be outer eyebrow raise up, AU5 be upper eyelid raise up with AU26 be lower jaw open etc..Each Facial action unit has instruction
Perfect corresponding SVM classifier.For example, may recognize that the part that interior eyebrow raises up is special by the interior eyebrow corresponding SVM classifier that raises up
Sign belongs to the probability value that interior eyebrow raises up, and is raised up the local feature that corresponding SVM classifier may recognize that outer eyebrow raises up by outer eyebrow
Belong to the probability value etc. that outer eyebrow raises up.
In the present embodiment, server identifies images to be recognized using preparatory trained micro- Expression Recognition model
When, face critical point detection and feature extraction etc. first can be carried out to each images to be recognized, to obtain the part of images to be recognized
Feature.Wherein, face key point algorithm can be but not limited to Ensemble of Regression Tress (abbreviation ERT) calculation
Method, SIFT (scale-invariant feature transform) algorithm, SURF (Speeded Up Robust
Features) algorithm, LBP (Local Binary Patterns) algorithm and HOG (Histogram of Oriented
Gridients) algorithm.Feature extraction algorithm can be calculated with CNN (Convolutional Neural Network, convolutional Neural net)
Method.The local feature is input in multiple SVM classifiers again, all parts by the input of multiple SVM classifiers pair are special
Sign is identified, is obtained the probability value corresponding with the Facial action unit of multiple SVM classifier outputs, probability value is greater than pre-
If the corresponding Facial action unit of the SVM classifier of threshold value is determined as target face motor unit.Wherein, target face movement is single
Member, which refers to, identifies images to be recognized according to micro- Expression Recognition model, the Facial action unit (Action got
Unit, AU).Probability value specifically can be the value between 0-1, if the probability value of output is 0.6, preset threshold 0.5, then generally
Rate value 0.6 is greater than preset threshold 0.5, then by 0.6 corresponding Facial action unit, the target face as images to be recognized is acted
Unit.Finally, accessed all target face motor units are carried out comprehensive assessment, obtains it and belong to micro- Expression Recognition mould
The corresponding probability of the preconfigured micro- expression type of type belongs to instantaneous confidence level of each identification expression type.It will be obtained
All target face motor units got carry out comprehensive assessment and specifically refer to the combination based on all target face motor units,
The probability that this combination belongs to preconfigured micro- expression type is obtained, to determine that it identifies the instantaneous confidence level of expression type.
S302: the maximum identification expression type of instantaneous confidence level is determined as the corresponding micro- expression type of images to be recognized.
Specifically, recognize each images to be recognized belong to it is at least one identification expression type instantaneous confidence level it
Afterwards, the maximum identification expression type of instantaneous confidence level need to be determined as to the corresponding micro- expression type of images to be recognized.For example, knowing
Being clipped to its images to be recognized to belong to the instantaneous confidence level of " love " this identification expression type is 0.9, and is belonged to " suspection " and " peaceful
It is quiet " the two identification expression types instantaneous confidence level be respectively 0.05, then by instantaneous confidence level be 0.9 corresponding identification expression
Type is determined as micro- expression type of the images to be recognized, to guarantee the accuracy of the micro- expression type identified.
S303: being based on micro- expression type queries mood attribute table of comparisons, obtains the corresponding mood attribute of images to be recognized.
Wherein, the mood attribute table of comparisons is pre-set for recording the corresponding mood attribute of each micro- expression type
Tables of data.In the mood attribute table of comparisons, it is divided into actively according to the corresponding mood enthusiasm degree of all micro- expression types
Mood, negative feeling and the neutral mood between active mood and negative feeling.Active mood refers to positive psychological state
Degree or the corresponding mood of state are the corresponding moods of benign, positive, the stable and constructive psychological condition of one kind, including but not
Be limited to like, happily, it is optimistic, trust, acceptable mood corresponding with micro- expression type such as pleasantly surprised.Negative feeling refers at certain
In concrete behavior, by external cause or the interior emotion for being unfavorable for continuing to complete work or normally think deeply generated by influence,
It is opposite with active mood, the corresponding feelings of micro- expression type such as including but not limited to detest, dislike, oppose, be discontented with, ignore and despise
Thread.Neutral mood is the mood between active mood and negative feeling, is the institute in addition to active mood and negative feeling
There is the corresponding mood of micro- expression type, mood corresponding with micro- expression type such as fatigue of including but not limited to cooling down.
In satisfaction evaluation method based on micro- expression provided by the present embodiment, first using micro- Expression Recognition model to every
One images to be recognized is identified, to determine that its corresponding at least one identifies the corresponding instantaneous confidence level of expression type, and is selected
The maximum identification expression type of instantaneous confidence level is taken to be determined as its corresponding micro- expression type, to guarantee the micro- expression identified
The accuracy of type.Be based on micro- expression type queries mood attribute table of comparisons again, can quick obtaining its corresponding mood attribute,
To ensure the acquisition efficiency of the mood attribute of images to be recognized.
In one embodiment, as shown in figure 4, step S205, that is, be based on each images to be recognized pair in target video segment
The mood attribute answered obtains the corresponding negative feeling probability of target video segment, specifically comprises the following steps:
S401: in statistics target video segment, the corresponding segment image sum of all images to be recognized.
In any target video segment, including benchmark image and timestamp when being preset after the timestamp of benchmark image
Between all images to be recognized in section.Since each benchmark image is an images to be recognized, i.e., in statistics target video segment,
The segment image sum of all images to be recognized refers in statistics target video segment that benchmark image and timestamp are in benchmark image
Timestamp after all images to be recognized in preset time period quantity sum.Since monitor terminal is in shooting video data
When, it is shot according to pre-set filming frequency, it is f, preset time period t that its filming frequency, which is arranged, then the target regards
In frequency segment, the segment image sum of all images to be recognized is N=t*f, with quick obtaining segment image sum.For example, pre-
The filming frequency f that monitor terminal is first arranged is 10 frames/second, if the shooting time t of a certain target video segment is 100 seconds, institute
The video image sum S=100*10=1000 of acquisition illustrates that captured target video segment includes 1000 frames figure to be identified
Picture.
S402: in statistics target video segment, mood attribute is the corresponding passive image of the images to be recognized of negative feeling
Quantity.
After determining a target video segment based on any benchmark image, trigger pre-set for counting passive figure
As the counter of quantity, the numerical value of its counter is made to be set as 1；Then, server is using micro- Expression Recognition model to reference map
Each images to be recognized as after is identified, the corresponding mood attribute of images to be recognized is obtained, if an images to be recognized
Mood attribute is negative feeling, then the numerical value of counter is made to add 1, until determining last frame figure to be identified in target video segment
After the mood attribute of picture, the numerical value of counter is determined as the corresponding passive amount of images of the target video segment.
S403: segment image sum and passive amount of images are calculated using negative feeling new probability formula, obtain mesh
The corresponding negative feeling probability of video clip is marked, negative feeling new probability formula is L=M/N, and L is negative feeling probability, and M is passiveness
Amount of images, N are segment image sum.
Specifically, server is after obtaining the corresponding segment image sum of target video segment and passive amount of images,
Pre-set negative feeling new probability formula can be used and quickly calculate its corresponding negative feeling probability.It is to be appreciated that institute is really
The negative feeling probability of fixed target video segment is bigger, illustrates target customer to the acquisition time of the target video segment (i.e.
In the timestamp of benchmark image to its preset time period), to receiving, the discontented probability of service is bigger, i.e., to seat personnel or
The topic point or service content that other business personnels currently talk about are more dissatisfied.
In satisfaction evaluation method based on micro- expression provided by the present embodiment, by counting each target video segment
Segment image sum and passive amount of images, then using negative feeling new probability formula to segment image sum and passive picture number
Amount is calculated, its corresponding negative feeling probability with quick obtaining, using the negative feeling probability as each mesh of overall merit
The basic unit that the corresponding target customer's satisfaction with the service of mark video clip is evaluated, can avoid accidental negative feeling to clothes
It is engaged in the interference of satisfaction evaluation, helps to improve the accuracy of the target of subsequent determination not full value.
In one embodiment, as shown in figure 5, after step S206, i.e., target video segment is being determined as passive feelings
After thread segment, the satisfaction evaluation method based on micro- expression further include:
S501: the corresponding passive number of fragments of negative feeling segment is updated.
Specifically, server can trigger pre-set for counting after it will determine first negative feeling segment
The counter of passive number of fragments makes the numerical value of its counter be set as 1.After determining first negative feeling segment, if
When server determines that either objective video clip is negative feeling segment, the numerical value of its counter can be made to add 1, to update passive feelings
The corresponding passive number of fragments of thread segment, realizes the real-time statistics to the negative feeling segment in video data.
S502: if passive number of fragments is greater than preset quantity threshold value, generating negative feeling prompting message, and by passive feelings
Thread prompting message is sent to terminal of attending a banquet.
Wherein, preset quantity threshold value is the pre-set amount threshold for being used to assess whether to need to send prompting message.
Specifically, server can compare the passiveness number of fragments in the corresponding passive number of fragments of update negative feeling segment every time
Whether preset quantity threshold value is greater than, then the passiveness number of fragments is greater than preset quantity threshold value, then automatically generates negative feeling prompting
Information, and the negative feeling prompting message is sent to terminal of attending a banquet so that the corresponding seat personnel of terminal of attending a banquet or other
Business personnel adjusts ditch call art in time, assists seat personnel or other business personnels are preferably target customer's service, with
Improve satisfaction of the target customer to the service received.
In satisfaction evaluation method based on micro- expression provided by the present embodiment, it is determined as disappearing by target video segment
After the mood segment of pole, the corresponding passive number of fragments of real-time update negative feeling segment is pre- to be greater than in passive number of fragments
If when amount threshold, to attending a banquet, terminal sends negative feeling prompting message, so that seat personnel or other business personnels are real-time
Know the current negative feeling state of target customer, carries out accordingly art in time and adjust, to help to improve target customer
Satisfaction to the service received.
In one embodiment, as shown in fig. 6, in step S207, i.e., using dissatisfaction formula to negative feeling segment into
Row processing obtains target not full value, specifically comprises the following steps:
S601: in statistics video data, the corresponding video image sum of all images to be recognized.
Specifically, in server statistics video data, the corresponding video image sum of all images to be recognized, in particular to
The quantity of all images to be recognized in video data is counted, and the quantity is determined as the corresponding video figure of all images to be recognized
As sum.Since monitor terminal is when shooting video data, is shot according to pre-set filming frequency, its shooting is set
Frequency is f, and server can be by the shooting time T of acquisition video data, can quick obtaining video data using formula S=T*f
In, the corresponding video image sum S of all images to be recognized.For example, preset monitor terminal filming frequency f be 10 frames/
Second, if the shooting time T of a certain video data is 600 seconds, acquired video image sum S=600*10=6000 is said
Bright captured video data includes 6000 frame images to be recognized.
S602: the timestamp query assessment coefficient vs table of the benchmark image according to each negative feeling segment obtains every
The corresponding metewand of one negative feeling segment.
Wherein, the metewand table of comparisons is pre-set for determining the number of the corresponding metewand of negative feeling segment
According to table.Timestamp (i.e. negative feeling segment in the metewand table of comparisons according to the corresponding benchmark image of negative feeling segment
First frame images to be recognized timestamp), judge that the timestamp in the position of the recording time of entire video data, determines it
Corresponding metewand.It, can also be with it is to be appreciated that the determination of the metewand can become larger with the variation of time
The variation of time gradually become smaller, mainly independently determined according to actual needs by user.For example, if the timestamp is in video counts
According to recording time before within 1/4, then assert its metewand K=1；The timestamp is in the recording time of video data
Before within 1/4-1/2, then assert its metewand K=1.1；The timestamp is in front of the recording time of video data
Within, then assert its metewand K=1.2；The timestamp is in 3/4-1 before the recording time of video data
Within, then assert its metewand K=1.3.The setting of metewand in this metewand table of comparisons, it is main to consider target visitor
Family is in the return visit or other service processes for receiving seat personnel or other business personnels offer, and the time is more long, Yong Huyue
It is easy impatient, it is easier to reflect target customer to the dissatisfaction for receiving service, therefore, according to negative feeling segment pair
The position of the timestamp for the benchmark image answered obtains corresponding metewand.
S603: obtaining in passive video clip, and mood attribute is the corresponding passive image of the images to be recognized of negative feeling
Quantity.
Since each passive video clip is the target video segment that negative feeling probability is greater than predetermined probabilities threshold value, and
In the calculating process (i.e. step S401-S403) of the negative feeling probability of target video segment, it has been computed to look over so as to check and has marked piece of video
Duan Zhong, mood attribute is the corresponding passive amount of images of the images to be recognized of negative feeling, as shown in step S402, therefore, step
The corresponding passive amount of images of target video segment can be directly determined as to the passive image of the passiveness video clip in rapid S603
Quantity, to avoid repeating.
S604: using dissatisfaction formula to video image sum, the corresponding metewand of each negative feeling segment and
Passive amount of images is calculated, and target not full value is obtained；Dissatisfaction formula isP is target not full value,
MiFor the passive amount of images of i-th of negative feeling segment, KiFor the metewand of i-th of negative feeling segment, n is passive feelings
The quantity of thread segment, S are video image sum.
For example, a certain video data has 10000 frame images to be recognized, there are the benchmark images of two negative feeling segments
Timestamp is respectively in 1/4 and last 1/4 before recording time, then the quantity n of negative feeling segment is 2；Wherein, first
There are the images to be recognized that 600 frame mood attributes are negative feeling in a negative feeling segment, passive amount of images is 600；
There are the images to be recognized that 800 mood attributes are negative feeling in second negative feeling segment, passive amount of images is
800；Then according to the dissatisfaction formulaIts target can quickly be calculated not
Full value.The calculating of this target not full value excludes the interference for the images to be recognized that the mood attribute accidentally occurred is negative feeling,
To guarantee the objectivity of the acquired target not dissatisfaction of full value reflection target customer, without to the collected feelings of institute
Thread attribute is that the images to be recognized of active mood and neutral mood is analyzed, and keeps its analytic process more targeted, calculates
Process is more simple and convenient.
In satisfaction evaluation method based on micro- expression provided by the present embodiment, by using dissatisfaction formula to view
Frequency total number of images, the corresponding metewand of each negative feeling segment and passive amount of images are calculated, can quick obtaining mesh
Not full value is marked, so that full value can not objectively respond the dissatisfaction of target customer to acquired target, excludes the feelings accidentally occurred
Thread attribute is the interference of the images to be recognized of negative feeling, without being active mood and neutral mood wait know to mood attribute
Other image is analyzed, and keeps its analytic process more targeted, calculating process is more simple and convenient.
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 satisfaction evaluation device based on micro- expression is provided, it should the satisfaction based on micro- expression
Satisfaction evaluation method in evaluating apparatus and above-described embodiment based on micro- expression corresponds.As shown in fig. 7, micro- table should be based on
The satisfaction evaluation device of feelings includes that video data obtains module 701, mood attribute obtains module 702, benchmark image determines mould
Block 703, target video segment form module 704, negative feeling probability obtains module 705, negative feeling segment determining module
706, full value does not obtain module 707 to target, target grading system obtains module 708 and benchmark image update module 709.Each function
Detailed description are as follows for module:
Video data obtains module 701, and for obtaining video data when target customer receives service, video data includes
An at least frame images to be recognized, the corresponding timestamp of each images to be recognized.
Mood attribute obtains module 702, for being identified using micro- Expression Recognition model to each images to be recognized, obtains
Take the corresponding mood attribute of images to be recognized.
Mood attribute is the first frame of negative feeling for the sequence according to timestamp by benchmark image determining module 703
Images to be recognized is determined as benchmark image.
Target video segment formed module 704, for based on timestamp after the timestamp of benchmark image preset time
Images to be recognized in section forms target video segment.
Negative feeling probability obtains module 705, for based on the corresponding feelings of images to be recognized each in target video segment
Thread attribute obtains the corresponding negative feeling probability of target video segment.
Negative feeling segment determining module 706, if being greater than for the corresponding negative feeling probability of target video segment default
Target video segment is then determined as negative feeling segment by probability threshold value, is based on negative feeling fragment update benchmark image, is repeated
It executes the images to be recognized based on timestamp after the timestamp of benchmark image in preset time period and forms target video segment.
Full value does not obtain module 707 to target, for being handled using dissatisfaction formula negative feeling segment, obtains
Target not full value.
Target grading system obtains module 708, for according to target, full value not to inquire grading system tables of data, acquisition target
Grading system.
Preferably, after negative feeling probability obtains module 705, the satisfaction evaluation device based on micro- expression is also wrapped
It includes:
Benchmark image update module 709, if general no more than default for the corresponding negative feeling probability of target video segment
Rate threshold value, then by after benchmark image, mood attribute is that the next frame images to be recognized of negative feeling is determined as benchmark image, weight
The multiple images to be recognized executed based on timestamp after the timestamp of benchmark image in preset time period forms target video piece
Section.
Preferably, mood attribute obtain module 702 include instantaneous confidence level acquiring unit, micro- expression type acquiring unit and
Mood attribute acquiring unit.
Instantaneous confidence level acquiring unit is obtained for being identified using micro- Expression Recognition model to each images to be recognized
Take the corresponding instantaneous confidence level of at least one identification expression type.
Micro- expression type acquiring unit, for the maximum identification expression type of instantaneous confidence level to be determined as images to be recognized
Corresponding micro- expression type.
Mood attribute acquiring unit obtains images to be recognized for being based on micro- expression type queries mood attribute table of comparisons
Corresponding mood attribute.
Preferably, it includes segment image sum statistic unit, passive amount of images system that negative feeling probability, which obtains module 705,
Count unit and negative feeling probability calculation unit.
Segment image sum statistic unit, for counting in target video segment, the corresponding segment of all images to be recognized
Total number of images.
Passive amount of images statistic unit, for counting in target video segment, mood attribute is negative feeling wait know
The corresponding passive amount of images of other image.
Negative feeling probability calculation unit, for using negative feeling new probability formula to segment image sum and passive image
Quantity is calculated, and the corresponding negative feeling probability of target video segment is obtained, and negative feeling new probability formula is L=M/N, and L is
Negative feeling probability, M are passive amount of images, and N is segment image sum.
Preferably, after negative feeling segment determining module 706, the satisfaction evaluation device based on micro- expression further includes
Passive number of fragments updating unit and negative feeling remind processing unit.
Passive number of fragments updating unit, for updating the corresponding passive number of fragments of negative feeling segment.
Negative feeling reminds processing unit, if being greater than preset quantity threshold value for passive number of fragments, generates passive feelings
Thread prompting message, and negative feeling prompting message is sent to terminal of attending a banquet.
Preferably, negative feeling segment determining module 706 includes the first benchmark image determination unit or the second reference map
As determination unit.
First benchmark image determination unit, for by after negative feeling segment, mood attribute to be the first of negative feeling
A images to be recognized is updated to benchmark image.
Second benchmark image determination unit, for by negative feeling segment, mood attribute to be last of negative feeling
A images to be recognized is updated to benchmark image.
Preferably, full value acquisition module 707 includes video image sum statistic unit, assessment system acquisition list to target
Member, passive amount of images acquiring unit and dissatisfaction computing unit.
Video image sum statistic unit, for counting in video data, the corresponding video image of all images to be recognized
Sum.
Assessment system acquiring unit, the timestamp query assessment system for the benchmark image according to each negative feeling segment
The number table of comparisons, obtains the corresponding metewand of each negative feeling segment.
Passive amount of images acquiring unit, for obtaining in passive video clip, mood attribute is negative feeling wait know
The corresponding passive amount of images of other image.
Dissatisfaction computing unit, for using dissatisfaction formula to video image sum, each negative feeling segment
Corresponding metewand and passive amount of images are calculated, and target not full value is obtained.Dissatisfaction formula isP is target not full value, MiFor the passive amount of images of i-th of negative feeling segment, KiFor i-th of passive feelings
The metewand of thread segment, S are video image sum.
Specific restriction about the satisfaction evaluation device based on micro- expression may refer to above for based on micro- expression
Satisfaction evaluation method restriction, details are not described herein.Each mould in the above-mentioned satisfaction evaluation device based on micro- expression
Block can be realized fully or partially through software, hardware and combinations thereof.Above-mentioned each module can be embedded in the form of hardware or independence
In processor in computer equipment, it can also be stored in a software form in the memory in computer equipment, in order to
Processor, which calls, executes the corresponding operation of the above modules.
In one embodiment, a kind of computer equipment is provided, which can be server, internal junction
Composition can be as shown in Figure 8.The computer equipment include by system bus connect processor, memory, network interface and
Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment
Include 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 execute the data for using or being formed during the satisfaction evaluation method based on micro- expression, such as video
Data.The network interface of the computer equipment is used to communicate with external terminal by network connection.The computer program is located
It manages when device executes to realize a kind of satisfaction evaluation method based on micro- expression.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, processor are realized in above-described embodiment when executing computer program based on micro- table
The satisfaction evaluation method of feelings, such as shown in S201-S209 or Fig. 3 to Fig. 6 shown in Fig. 2, to avoid repeating, here not
It repeats again.Alternatively, processor is realized in satisfaction evaluation device this embodiment based on micro- expression when executing computer program
Each module/unit function, such as the satisfaction evaluation device shown in Fig. 7 based on micro- expression include video data obtain mould
Block 701, mood attribute obtain module 702, benchmark image determining module 703, target video segment and form module 704, passive feelings
Thread probability obtains module 705, negative feeling segment determining module 706, target, and full value does not obtain module 707, target grading system
The function of module 708 and benchmark image update module 709 is obtained, to avoid repeating, which is not described herein again.
In one embodiment, a computer readable storage medium is provided, meter is stored on the computer readable storage medium
Calculation machine program, the computer program realize the satisfaction evaluation side based on micro- expression in above-described embodiment when being executed by processor
Method, such as shown in S201-S209 or Fig. 3 to Fig. 6 shown in Fig. 2, to avoid repeating, which is not described herein again.Alternatively, the meter
Calculation machine program realized when being executed by processor each module in above-mentioned satisfaction evaluation device this embodiment based on micro- expression/
The function of unit, such as the satisfaction evaluation device shown in Fig. 7 based on micro- expression include that video data obtains module 701, feelings
Thread attribute obtains module 702, benchmark image determining module 703, target video segment formation module 704, negative feeling probability and obtains
Full value does not obtain module 707, target grading system obtains module for modulus block 705, negative feeling segment determining module 706, target
708 and benchmark image update module 709 function, to avoid repeating, which is 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 satisfaction evaluation method based on micro- expression characterized by comprising
Video data when target customer receives service is obtained, the video data includes an at least frame images to be recognized, each
The corresponding timestamp of the images to be recognized；
Each images to be recognized is identified using micro- Expression Recognition model, obtains the corresponding feelings of the images to be recognized
Thread attribute；
According to the sequence of the timestamp, images to be recognized described in first frame of the mood attribute for negative feeling is determined as
Benchmark image；
Images to be recognized based on the timestamp after the timestamp of the benchmark image in preset time period forms target
Video clip；
Based on the corresponding mood attribute of the images to be recognized each in the target video segment, the target video piece is obtained
The corresponding negative feeling probability of section；
If the corresponding negative feeling probability of the target video segment is greater than predetermined probabilities threshold value, by the target video segment
It is determined as negative feeling segment, is based on the negative feeling fragment update benchmark image, repeats described based on the time
It stabs the images to be recognized after the timestamp of the benchmark image in preset time period and forms target video segment；
The negative feeling segment is handled using dissatisfaction formula, obtains target not full value；
According to the target, full value does not inquire grading system tables of data, obtains target grading system.
2. as described in claim 1 based on the satisfaction evaluation method of micro- expression, which is characterized in that be based on the mesh described
The corresponding mood attribute of each images to be recognized in video clip is marked, the corresponding passive feelings of the target video segment are obtained
After thread probability, the satisfaction evaluation method based on micro- expression further include:
If the corresponding negative feeling probability of the target video segment is not more than predetermined probabilities threshold value, by the benchmark image it
Afterwards, the mood attribute is that images to be recognized described in the next frame of negative feeling is determined as benchmark image, repeats the base
Images to be recognized after timestamp of the timestamp in the benchmark image in preset time period forms target video piece
Section.
3. as described in claim 1 based on the satisfaction evaluation method of micro- expression, which is characterized in that described to be known using micro- expression
Other model identifies each images to be recognized, obtains the corresponding mood attribute of the images to be recognized, comprising:
Each images to be recognized is identified using micro- Expression Recognition model, obtains at least one identification expression type pair
The instantaneous confidence level answered；
The instantaneous maximum identification expression type of confidence level is determined as the corresponding micro- expression type of the images to be recognized；
Based on micro- expression type queries mood attribute table of comparisons, the corresponding mood attribute of the images to be recognized is obtained.
4. as described in claim 1 based on the satisfaction evaluation method of micro- expression, which is characterized in that described to be based on the target
The corresponding mood attribute of each images to be recognized, obtains the corresponding negative feeling of the target video segment in video clip
Probability, comprising:
It counts in the target video segment, the corresponding segment image sum of all images to be recognized；
It counts in the target video segment, the mood attribute is the corresponding passive picture number of the images to be recognized of negative feeling
Amount；
The segment image sum and the passive amount of images are calculated using negative feeling new probability formula, described in acquisition
The corresponding negative feeling probability of target video segment, the negative feeling new probability formula are L=M/N, and L is that the negative feeling is general
Rate, M are the passive amount of images, and N is the segment image sum.
5. as described in claim 1 based on the satisfaction evaluation method of micro- expression, which is characterized in that described by the target
Video clip is determined as after negative feeling segment, the satisfaction evaluation method based on micro- expression further include:
Update the corresponding passive number of fragments of the negative feeling segment；
If the passiveness number of fragments is greater than preset quantity threshold value, negative feeling prompting message is generated, and by the passive feelings
Thread prompting message is sent to terminal of attending a banquet.
6. as described in claim 1 based on the satisfaction evaluation method of micro- expression, which is characterized in that described to be based on the passiveness
Mood fragment update benchmark image, comprising:
After the negative feeling segment, the mood attribute is that first images to be recognized of negative feeling is updated to
Benchmark image；Alternatively,
By in the negative feeling segment, the mood attribute is that the last one described images to be recognized of negative feeling is updated to
Benchmark image.
7. as described in claim 1 based on the satisfaction evaluation method of micro- expression, which is characterized in that described to use dissatisfaction
Formula handles the negative feeling segment, obtains target not full value, comprising:
It counts in the video data, the corresponding video image sum of all images to be recognized；
The timestamp query assessment coefficient vs table of benchmark image according to each negative feeling segment obtains each described
The corresponding metewand of negative feeling segment；
It obtains in the passive video clip, the mood attribute is the corresponding passive picture number of the images to be recognized of negative feeling
Amount；
To video image sum, the corresponding metewand of each negative feeling segment and disappeared using dissatisfaction formula
Pole amount of images is calculated, and target not full value is obtained；The dissatisfaction formula isP is discontented for target
Value, MiFor the passive amount of images of i-th of negative feeling segment, KiFor the assessment system of i-th of negative feeling segment
Number, n are the quantity of negative feeling segment, and S is video image sum.
8. a kind of satisfaction evaluation device based on micro- expression characterized by comprising
Video data obtains module, and for obtaining video data when target customer receives service, the video data includes extremely
A few frame images to be recognized, each corresponding timestamp of the images to be recognized；
Mood attribute obtains module, for being identified using micro- Expression Recognition model to each images to be recognized, obtains
The corresponding mood attribute of the images to be recognized；
The mood attribute is the first of negative feeling for the sequence according to the timestamp by benchmark image determining module
Images to be recognized described in frame is determined as benchmark image；
Target video segment formed module, for based on the timestamp after the timestamp of the benchmark image preset time
Images to be recognized in section forms target video segment；
Negative feeling probability obtains module, for based on the corresponding feelings of the images to be recognized each in the target video segment
Thread attribute obtains the corresponding negative feeling probability of the target video segment；
Negative feeling segment determining module, if being greater than predetermined probabilities for the corresponding negative feeling probability of the target video segment
The target video segment is then determined as negative feeling segment by threshold value, is based on the negative feeling fragment update benchmark image,
Repeat the images to be recognized based on the timestamp after the timestamp of the benchmark image in preset time period
Form target video segment；
Full value does not obtain module to target, for handling using dissatisfaction formula the negative feeling segment, obtains mesh
Mark not full value；
Target grading system obtains module, for according to the target, full value inquiry grading system tables of data, acquisition target not to be commented
Graduation.
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
Based on the satisfaction evaluation method of micro- expression described in 7 any one.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In satisfaction of the realization based on micro- expression as described in any one of claim 1 to 7 when the computer program is executed by processor
Evaluation method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043295.3A CN109858405A (en) | 2019-01-17 | 2019-01-17 | Satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043295.3A CN109858405A (en) | 2019-01-17 | 2019-01-17 | Satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109858405A true CN109858405A (en) | 2019-06-07 |
Family
ID=66895104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910043295.3A Pending CN109858405A (en) | 2019-01-17 | 2019-01-17 | Satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858405A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415015A (en) * | 2019-06-19 | 2019-11-05 | 深圳壹账通智能科技有限公司 | Product degree of recognition analysis method, device, terminal and computer readable storage medium |
| CN110647812A (en) * | 2019-08-19 | 2020-01-03 | 平安科技（深圳）有限公司 | Tumble behavior detection processing method and device, computer equipment and storage medium |
| CN110705349A (en) * | 2019-08-26 | 2020-01-17 | 深圳壹账通智能科技有限公司 | Customer satisfaction recognition method, device, terminal and medium based on micro expression |
| CN112347236A (en) * | 2020-11-16 | 2021-02-09 | 友谊国际工程咨询股份有限公司 | Intelligent engineering consultation method and system based on AI (Artificial Intelligence) calculated quantity and computer equipment thereof |
| CN112686195A (en) * | 2021-01-07 | 2021-04-20 | 风变科技（深圳）有限公司 | Emotion recognition method and device, computer equipment and storage medium |
| CN113822229A (en) * | 2021-10-28 | 2021-12-21 | 重庆科炬企业孵化器有限公司 | Expression recognition-oriented user experience evaluation modeling method and device |
| CN114445896A (en) * | 2022-01-28 | 2022-05-06 | 北京百度网讯科技有限公司 | Method and device for evaluating confidence degree of human statement content in video |
Citations (2)
| 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 |
| 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 |
-
2019
- 2019-01-17 CN CN201910043295.3A patent/CN109858405A/en active Pending
Patent Citations (2)
| 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 |
| 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 |
Non-Patent Citations (1)
| Title |
|---|
| 谢东亮;徐宇翔;: "基于人工智能的微表情识别技术", 科技与创新, no. 22, 30 November 2018 (2018-11-30), pages 35 - 37 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415015A (en) * | 2019-06-19 | 2019-11-05 | 深圳壹账通智能科技有限公司 | Product degree of recognition analysis method, device, terminal and computer readable storage medium |
| CN110647812A (en) * | 2019-08-19 | 2020-01-03 | 平安科技（深圳）有限公司 | Tumble behavior detection processing method and device, computer equipment and storage medium |
| CN110647812B (en) * | 2019-08-19 | 2023-09-19 | 平安科技（深圳）有限公司 | Tumble behavior detection processing method and device, computer equipment and storage medium |
| CN110705349A (en) * | 2019-08-26 | 2020-01-17 | 深圳壹账通智能科技有限公司 | Customer satisfaction recognition method, device, terminal and medium based on micro expression |
| CN112347236A (en) * | 2020-11-16 | 2021-02-09 | 友谊国际工程咨询股份有限公司 | Intelligent engineering consultation method and system based on AI (Artificial Intelligence) calculated quantity and computer equipment thereof |
| CN112686195A (en) * | 2021-01-07 | 2021-04-20 | 风变科技（深圳）有限公司 | Emotion recognition method and device, computer equipment and storage medium |
| CN113822229A (en) * | 2021-10-28 | 2021-12-21 | 重庆科炬企业孵化器有限公司 | Expression recognition-oriented user experience evaluation modeling method and device |
| CN114445896A (en) * | 2022-01-28 | 2022-05-06 | 北京百度网讯科技有限公司 | Method and device for evaluating confidence degree of human statement content in video |
| CN114445896B (en) * | 2022-01-28 | 2024-04-05 | 北京百度网讯科技有限公司 | Method and device for evaluating confidence of content of person statement in video |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109858405A (en) | Satisfaction evaluation method, apparatus, equipment and storage medium based on micro- expression | |
| WO2020253358A1 (en) | Service data risk control analysis processing method, apparatus and computer device | |
| CN109858958A (en) | Aim client orientation method, apparatus, equipment and storage medium based on micro- expression | |
| US10657651B2 (en) | Systems and methods for detection of significant and attractive components in digital images | |
| WO2020143156A1 (en) | Hotspot video annotation processing method and apparatus, computer device and storage medium | |
| JP2019509556A (en) | User credit evaluation method, apparatus and storage medium | |
| WO2021012790A1 (en) | Page data generation method and apparatus, computer device, and storage medium | |
| WO2020015104A1 (en) | Method, apparatus, computer device, and storage medium for predicting flow rate of passengers presenting security risk | |
| KR20200098875A (en) | System and method for providing 3D face recognition | |
| US9147128B1 (en) | Machine learning enhanced facial recognition | |
| CN110245132A (en) | Data exception detection method, device, computer readable storage medium and computer equipment | |
| CN108469975B (en) | Control display method and device, computer equipment and storage medium | |
| CN109977651A (en) | Man-machine recognition methods, device and electronic equipment based on sliding trace | |
| CN109063984B (en) | Method, apparatus, computer device and storage medium for risky travelers | |
| CN109376598A (en) | Facial expression image processing method, device, computer equipment and storage medium | |
| CN111192153B (en) | Crowd relation network construction method, device, computer equipment and storage medium | |
| CN110287103A (en) | Software product assessment processing method, device, computer equipment and storage medium | |
| CN110119877A (en) | A kind of choosing method and equipment of target employee | |
| WO2021012482A1 (en) | Method and device for generating group interest tag, computer device, and storage medium | |
| WO2023024670A1 (en) | Device clustering method and apparatus, and computer device and storage medium | |
| CN111079553A (en) | Method, device, equipment and storage medium for acquiring customer analysis data | |
| CN112131277A (en) | Medical data anomaly analysis method and device based on big data and computer equipment | |
| CN112819611A (en) | Fraud identification method, device, electronic equipment and computer-readable storage medium | |
| AU2019101158A4 (en) | A method of analyzing customer churn of credit cards by using logistics regression | |
| CN109697421A (en) | Evaluation method, device, computer equipment and storage medium based on micro- expression |
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 |