CN110287912A - Method, apparatus and medium are determined based on the target object affective state of deep learning - Google Patents
Method, apparatus and medium are determined based on the target object affective state of deep learning Download PDFInfo
- Publication number
- CN110287912A CN110287912A CN201910576185.3A CN201910576185A CN110287912A CN 110287912 A CN110287912 A CN 110287912A CN 201910576185 A CN201910576185 A CN 201910576185A CN 110287912 A CN110287912 A CN 110287912A
- Authority
- CN
- China
- Prior art keywords
- affective state
- target object
- model
- face
- deep learning
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- 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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Image Analysis (AREA)
Abstract
This application discloses a kind of target object affective states based on deep learning to determine method, comprising: includes the picture frame of target object in acquisition video clip;Picture frame is separately input into Face datection model and behavioral value model, determines corresponding expression affective state and behavior affective state using the face characteristic and behavioural characteristic that identify;The target affective state of target object is determined according to expression affective state and behavior affective state.As it can be seen that compared to the prior art, the target object affective state provided in this embodiment based on deep learning determines that method further contemplates influence of the behavioural characteristic of target object to determining affective state, therefore can more accurately obtain the affective state of target object.Disclosed herein as well is a kind of target object affective state determining device and computer readable storage medium based on deep learning, all have above-mentioned beneficial effect.
Description
Technical field
The present invention relates to field of image recognition, in particular to a kind of target object affective state based on deep learning determines
Method, apparatus and computer readable storage medium.
Background technique
With the development of information technology, video technique in people's daily life using more more and more universal.For example, people
Can be used Internet chatroom carry out Video chat, trans-corporation by network progress video conference, subway, square, supermarket
Equal public places carry out video monitoring etc. by camera.Currently, in order to further such that people can be in the mistake of viewing video
Cheng Zhonggeng directly and accurately understands the affective state of the target object in video clip, improves the efficiency linked up between men,
Or timely early warning can be carried out to the target object of abnormal feeling, dangerous generation is avoided, is proposed a kind of based on depth
The target object affective state of habit determines method.In this method, by carrying out figure to from the picture frame extracted in video clip
As identification, the affective state of target object is determined according to the expression of the target object identified.But due to target object
Expression and actual affective state are not complete corresponding relationship, therefore determine target pair according to method in the prior art
Error will be present in the affective state of elephant.
Therefore, the accuracy for determining the affective state of target object how is improved, is that those skilled in the art need at present
Technical problems to be solved.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of target object affective state determination sides based on deep learning
Method can be improved the accuracy for determining the affective state of target object;It is a further object of the present invention to provide one kind based on deep
The target object affective state determining device and computer readable storage medium for spending study, all have above-mentioned beneficial effect.
In order to solve the above technical problems, the present invention provides a kind of target object affective state determination side based on deep learning
Method, comprising:
It include the picture frame of target object in acquisition video clip;
Described image frame is separately input into Face datection model and behavioral value model, it is special using the face identified
Behavioural characteristic of seeking peace determines corresponding expression affective state and behavior affective state;Wherein, the Face datection model and institute
Stating behavioral value model is the model being arranged by deep learning;
The target emotion shape of the target object is determined according to the expression affective state and the behavior affective state
State.
Preferably, described that described image frame is separately input into Face datection model and behavioral value model, utilize knowledge
Not Chu face characteristic and behavioural characteristic determine the process of corresponding expression affective state and behavior affective state, it is specific to wrap
It includes:
Described image frame is input in the Face datection model, the extraction of Face datection and facial key point is carried out,
Obtain the face coordinate of the target object;
The corresponding face image of the face coordinate is input in the facial expression disaggregated model trained in advance, is obtained
Corresponding expression affective state;
Described image frame is input in the behavioral value model, the extraction of human testing and human body key point is carried out,
Obtain the limbs coordinate of the target object;
The described image frame for being provided with the limbs coordinate is input in the limbs expression classification model trained in advance,
Obtain corresponding behavior affective state.
Preferably, it is described acquisition video clip in include target object picture frame after, further comprise:
The similarity between any two described image frame is calculated using the feature vector in each described image frame and is adopted
Sample distance threshold;
Initial clustering is carried out to each described image frame using the sampled distance threshold value and the similarity, and optimizes and obtains
Target cluster centre;
Key frame is determined using the target cluster centre;
It is corresponding, it is described that described image frame is separately input into Face datection model and behavioral value model, utilize knowledge
Not Chu face characteristic and behavioural characteristic determine the process of corresponding expression affective state and behavior affective state, specifically:
The key frame is separately input into Face datection model and behavioral value model, it is special using the face identified
Behavioural characteristic of seeking peace determines corresponding expression affective state and behavior affective state.
Preferably, the Face datection model and the behavioral value model are specially in Two-pathway CNN network
Face datection model and behavioral value model.
Preferably, further comprise:
Record the target object affective state and the corresponding time for determining the affective state.
In order to solve the above technical problems, the present invention also provides a kind of target object affective states based on deep learning to determine
Device, comprising:
Obtain module, for obtain include in video clip target object picture frame;
Identification module is utilized for described image frame to be separately input into Face datection model and behavioral value model
The face characteristic and behavioural characteristic identified determines corresponding expression affective state and behavior affective state;Wherein, the people
Face detection model and the behavioral value model are the model being arranged by deep learning;
Determining module, for determining the target object according to the expression affective state and the behavior affective state
Target affective state.
Preferably, further comprise:
Computing module, for calculating the phase between any two picture frame using the feature vector in each described image frame
Like degree and sampled distance threshold value;
Cluster module, for initially being gathered using the sampled distance threshold value and the similarity to each described image frame
Class, and optimize and obtain target cluster centre;
Determining module, for determining key frame using the target cluster centre.
In order to solve the above technical problems, the target object affective state the present invention also provides another kind based on deep learning is true
Determine device, comprising:
Memory, for storing computer program;
Processor realizes any of the above-described kind of target object feelings based on deep learning when for executing the computer program
The step of sense state determines method.
In order to solve the above technical problems, the present invention also provides a kind of computer readable storage medium, it is described computer-readable
Computer program is stored on storage medium, the computer program realizes that any of the above-described kind is based on depth when being executed by processor
The step of target object affective state of study determines method.
Target object affective state provided by the invention based on deep learning determines method, further contemplates target pair
The behavioural characteristic of elephant to determine affective state influence, by obtain video clip in include target object picture frame it
Afterwards, picture frame is separately input into Face datection model and behavioral value model, utilizes the face characteristic and behavior identified
Feature determines corresponding expression affective state and behavior affective state, then according to expression affective state and behavior affective state
It determines the target affective state of target object, therefore can more accurately obtain the affective state of target object.
In order to solve the above technical problems, the present invention also provides a kind of, the target object affective state based on deep learning is true
Determine device and computer readable storage medium, all has above-mentioned beneficial effect.
Detailed description of the invention
It in order to illustrate the embodiments of the present invention more clearly or the technical solution of the prior art, below will be to embodiment or existing
Attached drawing needed in technical description is briefly described, it should be apparent that, the accompanying drawings in the following description is only this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to the attached drawing of offer.
Fig. 1 is a kind of stream that method is determined based on the target object affective state of deep learning provided in an embodiment of the present invention
Cheng Tu;
Fig. 2 is a kind of knot of the target object affective state determining device based on deep learning provided in an embodiment of the present invention
Composition;
Fig. 3 is another target object affective state determining device based on deep learning provided in an embodiment of the present invention
Structure chart.
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 only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The core of the embodiment of the present invention is to provide a kind of target object affective state based on deep learning and determines method, energy
It is enough to improve the accuracy for determining the affective state of target object;Another core of the invention is to provide a kind of based on deep learning
Target object affective state determining device and computer readable storage medium, all have above-mentioned beneficial effect.
It is right with reference to the accompanying drawings and detailed description in order to make those skilled in the art more fully understand the present invention program
The present invention is described in further detail.
Fig. 1 is a kind of stream that method is determined based on the target object affective state of deep learning provided in an embodiment of the present invention
Cheng Tu.As shown in Figure 1, a kind of target object affective state based on deep learning determines that method includes:
S10: including the picture frame of target object in acquisition video clip.
Specifically, the present embodiment is further to extract in video clip after getting video clip for true first
Set the goal object affective state picture frame, and due to needing to carry out analysis identification to picture frame frame, the picture frame
In need to include target object.
S20: picture frame is separately input into Face datection model and behavioral value model, special using the face identified
Behavioural characteristic of seeking peace determines corresponding expression affective state and behavior affective state;Wherein, Face datection model and behavior inspection
Surveying model is the model being arranged by deep learning.
Specifically, picture frame is then separately input into Face datection model and behavioral value after getting picture frame
In model, recognition of face and Activity recognition are carried out, the face characteristic of target object is obtained by recognition of face, according to Activity recognition
It obtains the behavioural characteristic of target object, and corresponding expression affective state and row is obtained according to face characteristic and behavioural characteristic respectively
For affective state.
S30: the target affective state of target object is determined according to expression affective state and behavior affective state.
It is understood that the emotion contained under different behavior states is different for same expression.Example
Such as, for this affective state of cryying, if not combining specific behavior, we can not judge that this people is the tears that weep through sorrow
Still it is so happy as to weep.Specifically, analyzed in the present embodiment further combined with expression affective state and behavior affective state,
Such as can be weighted by pre-set expression affective state and the corresponding weight of behavior affective state, it determines
The target affective state of target object out.
It should be noted that in practical applications, can be carried out according to the target affective state for the target object determined
Corresponding operation: for example, it is assumed that be the video clip obtained from the camera for being set to public place, it can be by identifying
The target affective state of target object in video clip, obtains the target object in abnormal object affective state, such as angry,
Indignation etc., so as to carry out early warning preparation, in advance so as to control in time abnormal conditions.
As it can be seen that compared to the prior art, the target object affective state provided in this embodiment based on deep learning determines
Method further contemplates influence of the behavioural characteristic of target object to affective state is determined, by wrapping in obtaining video clip
After the picture frame for including target object, picture frame is separately input into Face datection model and behavioral value model, is utilized
The face characteristic and behavioural characteristic identified determines corresponding expression affective state and behavior affective state, then according to expression
Affective state and behavior affective state determine the target affective state of target object, therefore can more accurately obtain target object
Affective state.
On the basis of the above embodiments, the present embodiment has made further instruction and optimization to technical solution, specifically,
Picture frame is separately input into Face datection model and behavioral value model, the face characteristic and behavioural characteristic identified is utilized
The process for determining corresponding expression affective state and behavior affective state, specifically includes:
Picture frame is input in Face datection model, the extraction of Face datection and facial key point is carried out, obtains target
The face coordinate of object;
The corresponding face image of face coordinate is input in the facial expression disaggregated model trained in advance, obtains correspondence
Expression affective state;
Picture frame is input in behavioral value model, the extraction of human testing and human body key point is carried out, obtains target
The limbs coordinate of object;
The picture frame for being provided with limbs coordinate is input in the limbs expression classification model trained in advance, obtains correspondence
Behavior affective state.
Specifically, picture frame is separately input into Face datection model and behavioral value model first, face inspection is carried out
Survey and the extraction of facial key point and the extraction for carrying out human testing and human body key point, obtain target object face coordinate and
The limbs coordinate of target object.
Then, the corresponding face image of face coordinate is input in the facial expression disaggregated model trained in advance, is obtained
Corresponding expression affective state out;The picture frame for being provided with limbs coordinate is input to the limbs expression classification mould trained in advance
In type, corresponding behavior affective state is obtained.
In the present embodiment, Face datection model and behavioral value model are specially in Two-pathway CNN network
Face datection model and behavioral value model.
Specifically, the present embodiment be Face datection and human testing are carried out by Two-pathway CNN network, thus
Opposite detection efficiency can be improved, save detection time.Also, in Two-pathway CNN network, Face datection model can
To be specially Mt-CNN convolutional neural networks, behavioral value model can be specially faster-RCNN convolutional neural networks, this reality
Example is applied not limit this.In addition, facial expression disaggregated model can be specifically local CNN neural network, limbs expression point
Class model can be specifically global CNN neural network, and the present embodiment does not do specific restriction to this yet.
As it can be seen that the present embodiment is to utilize model trained in advance, and obtain expression affective state and behavior according to picture frame
Affective state, therefore can be derived that more accurate recognition result.
It is understood that video clip is as a kind of unstructured dynamic number being made of continuous associated images frame
According to often comprising some redundancies.Such as in the same camera lens, the picture frame and the image at t+1 moment of video t moment
Often therefore difference very little carries out the emotion recognition of target object in foundation video clip, really to frame on vision content and feature
During the affective state for making target object, picture frame all in video clip is all used for the emotion shape of target object
The analysis of state, will be so that analytic process be complicated, and calculation amount is huge, and operates redundancy.Therefore, the present embodiment is in above-mentioned implementation
On the basis of example, further instruction and optimization are made to technical solution, specifically, including target in obtaining video clip
After the picture frame of object, further comprise:
The similarity and sampled distance threshold between any two picture frame are calculated using the feature vector in each picture frame
Value;
Initial clustering is carried out to each picture frame using sampled distance threshold value and similarity, and optimizes and obtains in target cluster
The heart;
Key frame is determined using target cluster centre.
Specifically, being the video clip that acquisition includes target object first, then extracting each video in the present embodiment
Each picture frame in segment obtains the feature vector in each picture frame, calculates the Euclidean distance between any two picture frame,
Similarity i.e. between any two picture frame, recycles calculated Euclidean distance to calculate sampled distance threshold value.
More specifically, the calculation method of the Euclidean distance between any two picture frame is as follows:
Wherein, dis (Fi,Fj) indicate any two picture frame feature vector Euclidean distance;FiAnd FjIndicate video figure
The feature vector of any two frames i and j as in;R indicates each picture frame;N is the dimension of the feature vector of picture frame.
Specifically, the calculation method of sampled distance threshold value is as follows:
Wherein, c constant, N are the quantity of picture frame;Indicate the number of the Euclidean distance calculated.
Different sampled distance threshold values can be chosen to different video clips by this method, also, calculated
The initial clustering that sampled distance threshold value less than normal can make subsequent acquisition more is conducive to the class heart and merges and secondary cluster.
After calculating sampled distance threshold value and similarity, using sampled distance threshold value and similarity in video clip
Picture frame carry out initial clustering, obtain initial cluster center set C and clusters number K.
Specifically, by random selection one picture frame feature vector be used as cluster centre, calculating it is remaining each
Feature vector is at a distance from cluster centre;Then judge whether the minimum value in each distance is less than sampled distance threshold value, that is, judge
Min(dis)≥2*d;If so, cluster data is added 1, and using the picture frame as a cluster centre, that is, by the picture frame
Corresponding feature vector is added in the set of cluster centre;If it is not, then the corresponding feature vector of the picture frame is deleted;Again from
Next picture frame cluster centre the most is chosen in picture frame and carries out judgement calculating, until by all picture frames of video clip
It has carried out judgement to calculate, has obtained initial cluster center set C and clusters number K.
Then, the initial cluster center set C obtained is clustered using K-means algorithm again, is updated
Cluster centre set G successively calculates the distance between two cluster centres in cluster centre set G, if Min (dis)≤2*d,
The two cluster centres are then merged by Sequence cluster, and calculate corresponding updated cluster centre;Otherwise, it chooses another
Group cluster center carries out judgement calculating, until each cluster centre in traversal cluster centre set G, thus the mesh after being optimized
Mark cluster centre.
After obtaining target cluster centre, the image nearest from each target cluster centre is chosen from all picture frames
Key frame is separately input into Face datection model and behavioral value model by key frame of the frame as the video clip, correspondence,
Corresponding expression affective state and behavior affective state are determined using the face characteristic and behavioural characteristic that identify.
In the present embodiment, by the key frame for further selecting the video object, it is possible to reduce image interframe it is a large amount of superfluous
Remaining information, the information that more condensed compactly expression video clip includes, the emotion of target object is carried out using the key frame selected
The determination of state can be further reduced calculation amount, improve the real-time for determining the affective state of target object.
On the basis of the above embodiments, the present embodiment has made further instruction and optimization to technical solution, specifically,
The present embodiment further comprises:
The affective state of record target object and the corresponding time for determining affective state.
Specifically, the target affective state that target object is determined according to expression affective state and behavior affective state it
Afterwards, record target object affective state and the corresponding time for determining affective state.It should be noted that specific record
Mode can be with text or record in table form, and the present embodiment does not limit this.More specifically, it can be with memory
The modes such as item, hard disk, TF (Trans-flash Card) card and SD (Secure Digital Memory Card) card are deposited
Storage, is selected, the present embodiment is not limited this with specific reference to actual demand.
In the present embodiment, affective state and the corresponding time for determining affective state by record target object, energy
Enough convenient for the emotion situation of change of subsequent further analysis target object, so as to further promote usage experience.
Above for a kind of implementation for determining method based on the target object affective state of deep learning provided by the invention
Example is described in detail, and the present invention also provides a kind of target object emotions based on deep learning corresponding with this method
State determination device and computer readable storage medium, due to the embodiment and side of device and computer readable storage medium part
The embodiment of method part mutually correlates, therefore the embodiment of device and computer readable storage medium part refers to method part
Embodiment description, wouldn't repeat here.
Fig. 2 is a kind of knot of the target object affective state determining device based on deep learning provided in an embodiment of the present invention
Composition, as shown in Fig. 2, a kind of target object affective state determining device based on deep learning includes:
Obtain module 21, for obtain include in video clip target object picture frame;
Identification module 22 utilizes knowledge for picture frame to be separately input into Face datection model and behavioral value model
Not Chu face characteristic and behavioural characteristic determine corresponding expression affective state and behavior affective state;Wherein, Face datection
Model and behavioral value model are the model being arranged by deep learning;
Determining module 23, for determining the target emotion of target object according to expression affective state and behavior affective state
State.
Target object affective state determining device provided in an embodiment of the present invention based on deep learning has above-mentioned be based on
The target object affective state of deep learning determines the beneficial effect of method.
As preferred embodiment, further comprise:
Computing module, for calculating the similarity between any two picture frame using the feature vector in each picture frame
With sampled distance threshold value;
Cluster module for carrying out initial clustering to each picture frame using sampled distance threshold value and similarity, and optimizes
To target cluster centre;
Determining module, for determining key frame using target cluster centre.
Fig. 3 is another target object affective state determining device based on deep learning provided in an embodiment of the present invention
Structure chart, as shown in figure 3, a kind of target object affective state determining device based on deep learning includes:
Memory 31, for storing computer program;
Processor 32 realizes such as the above-mentioned target object affective state based on deep learning when for executing computer program
The step of determining method.
Target object affective state determining device provided in an embodiment of the present invention based on deep learning has above-mentioned be based on
The target object affective state of deep learning determines the beneficial effect of method.
In order to solve the above technical problems, the present invention also provides a kind of computer readable storage medium, computer-readable storage
It is stored with computer program on medium, such as the above-mentioned target pair based on deep learning is realized when computer program is executed by processor
The step of determining method as affective state.
Computer readable storage medium provided in an embodiment of the present invention has the above-mentioned target object feelings based on deep learning
Sense state determines the beneficial effect of method.
Method, apparatus and calculating are determined to the target object affective state provided by the present invention based on deep learning above
Machine readable storage medium storing program for executing is described in detail.Specific embodiment used herein to the principle of the present invention and embodiment into
Elaboration is gone, the above description of the embodiment is only used to help understand the method for the present invention and its core ideas.It should be pointed out that pair
For those skilled in the art, without departing from the principle of the present invention, the present invention can also be carried out
Some improvements and modifications, these improvements and modifications also fall within the scope of protection of the claims of the present invention.
Each embodiment is described in a progressive manner in specification, the highlights of each of the examples are with other realities
The difference of example is applied, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment
Speech, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is referring to method part illustration
?.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Profession
Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered
Think beyond the scope of this invention.
Claims (9)
1. a kind of target object affective state based on deep learning determines method characterized by comprising
It include the picture frame of target object in acquisition video clip;
Described image frame is separately input into Face datection model and behavioral value model, using the face characteristic that identifies and
Behavioural characteristic determines corresponding expression affective state and behavior affective state;Wherein, the Face datection model and the row
It is the model being arranged by deep learning for detection model;
The target affective state of the target object is determined according to the expression affective state and the behavior affective state.
2. the method according to claim 1, wherein described be separately input into Face datection mould for described image frame
In type and behavioral value model, corresponding expression affective state and row are determined using the face characteristic and behavioural characteristic that identify
For the process of affective state, specifically include:
Described image frame is input in the Face datection model, the extraction of Face datection and facial key point is carried out, obtains
The face coordinate of the target object;
The corresponding face image of the face coordinate is input in the facial expression disaggregated model trained in advance, obtains correspondence
Expression affective state;
Described image frame is input in the behavioral value model, the extraction of human testing and human body key point is carried out, obtains
The limbs coordinate of the target object;
The described image frame for being provided with the limbs coordinate is input in the limbs expression classification model trained in advance, is obtained
Corresponding behavior affective state.
3. the method according to claim 1, wherein including target object in the acquisition video clip
After picture frame, further comprise:
Using the feature vector in each described image frame calculate similarity between any two described image frame and sampling away from
From threshold value;
Initial clustering is carried out to each described image frame using the sampled distance threshold value and the similarity, and optimizes and obtains target
Cluster centre;
Key frame is determined using the target cluster centre;
It is corresponding, it is described that described image frame is separately input into Face datection model and behavioral value model, using identifying
Face characteristic and behavioural characteristic determine the process of corresponding expression affective state and behavior affective state, specifically:
The key frame is separately input into Face datection model and behavioral value model, using the face characteristic that identifies and
Behavioural characteristic determines corresponding expression affective state and behavior affective state.
4. the method according to claim 1, wherein the Face datection model and behavioral value model tool
Body is the Face datection model and behavioral value model in Two-pathway CNN network.
5. method according to any one of claims 1 to 4, which is characterized in that further comprise:
Record the target object affective state and the corresponding time for determining the affective state.
6. a kind of target object affective state determining device based on deep learning characterized by comprising
Obtain module, for obtain include in video clip target object picture frame;
Identification module utilizes identification for described image frame to be separately input into Face datection model and behavioral value model
Face characteristic and behavioural characteristic out determines corresponding expression affective state and behavior affective state;Wherein, the face inspection
It surveys model and the behavioral value model is the model being arranged by deep learning;
Determining module, for determining the mesh of the target object according to the expression affective state and the behavior affective state
Mark affective state.
7. device according to claim 6, which is characterized in that further comprise:
Computing module, for calculating the similarity between any two picture frame using the feature vector in each described image frame
With sampled distance threshold value;
Cluster module, for carrying out initial clustering to each described image frame using the sampled distance threshold value and the similarity,
And optimizes and obtain target cluster centre;
Determining module, for determining key frame using the target cluster centre.
8. a kind of target object affective state determining device based on deep learning characterized by comprising
Memory, for storing computer program;
Processor is realized when for executing the computer program and is based on deep learning as described in any one of claim 1 to 5
Target object affective state the step of determining method.
9. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program, when the computer program is executed by processor realize as it is described in any one of claim 1 to 5 based on deep learning
The step of target object affective state determines method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910576185.3A CN110287912A (en) | 2019-06-28 | 2019-06-28 | Method, apparatus and medium are determined based on the target object affective state of deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910576185.3A CN110287912A (en) | 2019-06-28 | 2019-06-28 | Method, apparatus and medium are determined based on the target object affective state of deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110287912A true CN110287912A (en) | 2019-09-27 |
Family
ID=68019647
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910576185.3A Pending CN110287912A (en) | 2019-06-28 | 2019-06-28 | Method, apparatus and medium are determined based on the target object affective state of deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110287912A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111882625A (en) * | 2020-07-07 | 2020-11-03 | 北京达佳互联信息技术有限公司 | Method and device for generating dynamic graph, electronic equipment and storage medium |
CN113723165A (en) * | 2021-03-25 | 2021-11-30 | 山东大学 | Method and system for detecting dangerous expressions of people to be detected based on deep learning |
CN114064969A (en) * | 2021-11-19 | 2022-02-18 | 浙江大学 | Dynamic picture linkage display device based on emotional curve |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106295568A (en) * | 2016-08-11 | 2017-01-04 | 上海电力学院 | The mankind's naturalness emotion identification method combined based on expression and behavior bimodal |
CN106503646A (en) * | 2016-10-19 | 2017-03-15 | 竹间智能科技(上海)有限公司 | Multi-modal emotion identification system and method |
CN106851437A (en) * | 2017-01-17 | 2017-06-13 | 南通同洲电子有限责任公司 | A kind of method for extracting video frequency abstract |
CN107220585A (en) * | 2017-03-31 | 2017-09-29 | 南京邮电大学 | A kind of video key frame extracting method based on multiple features fusion clustering shots |
CN107808146A (en) * | 2017-11-17 | 2018-03-16 | 北京师范大学 | A kind of multi-modal emotion recognition sorting technique |
CN108520250A (en) * | 2018-04-19 | 2018-09-11 | 北京工业大学 | A kind of human motion sequence extraction method of key frame |
CN109145754A (en) * | 2018-07-23 | 2019-01-04 | 上海电力学院 | Merge the Emotion identification method of facial expression and limb action three-dimensional feature |
CN109766759A (en) * | 2018-12-12 | 2019-05-17 | 成都云天励飞技术有限公司 | Emotion identification method and Related product |
-
2019
- 2019-06-28 CN CN201910576185.3A patent/CN110287912A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106295568A (en) * | 2016-08-11 | 2017-01-04 | 上海电力学院 | The mankind's naturalness emotion identification method combined based on expression and behavior bimodal |
CN106503646A (en) * | 2016-10-19 | 2017-03-15 | 竹间智能科技(上海)有限公司 | Multi-modal emotion identification system and method |
CN106851437A (en) * | 2017-01-17 | 2017-06-13 | 南通同洲电子有限责任公司 | A kind of method for extracting video frequency abstract |
CN107220585A (en) * | 2017-03-31 | 2017-09-29 | 南京邮电大学 | A kind of video key frame extracting method based on multiple features fusion clustering shots |
CN107808146A (en) * | 2017-11-17 | 2018-03-16 | 北京师范大学 | A kind of multi-modal emotion recognition sorting technique |
CN108520250A (en) * | 2018-04-19 | 2018-09-11 | 北京工业大学 | A kind of human motion sequence extraction method of key frame |
CN109145754A (en) * | 2018-07-23 | 2019-01-04 | 上海电力学院 | Merge the Emotion identification method of facial expression and limb action three-dimensional feature |
CN109766759A (en) * | 2018-12-12 | 2019-05-17 | 成都云天励飞技术有限公司 | Emotion identification method and Related product |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111882625A (en) * | 2020-07-07 | 2020-11-03 | 北京达佳互联信息技术有限公司 | Method and device for generating dynamic graph, electronic equipment and storage medium |
CN111882625B (en) * | 2020-07-07 | 2024-04-05 | 北京达佳互联信息技术有限公司 | Method, device, electronic equipment and storage medium for generating dynamic diagram |
CN113723165A (en) * | 2021-03-25 | 2021-11-30 | 山东大学 | Method and system for detecting dangerous expressions of people to be detected based on deep learning |
CN113723165B (en) * | 2021-03-25 | 2022-06-07 | 山东大学 | Method and system for detecting dangerous expressions of people to be detected based on deep learning |
CN114064969A (en) * | 2021-11-19 | 2022-02-18 | 浙江大学 | Dynamic picture linkage display device based on emotional curve |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Fan et al. | Lasot: A high-quality benchmark for large-scale single object tracking | |
CN110675475B (en) | Face model generation method, device, equipment and storage medium | |
US20230049135A1 (en) | Deep learning-based video editing method, related device, and storage medium | |
CN107423398A (en) | Exchange method, device, storage medium and computer equipment | |
CN108229268A (en) | Expression Recognition and convolutional neural networks model training method, device and electronic equipment | |
CN110287912A (en) | Method, apparatus and medium are determined based on the target object affective state of deep learning | |
CN111881776B (en) | Dynamic expression acquisition method and device, storage medium and electronic equipment | |
CN113822254B (en) | Model training method and related device | |
CN110503076A (en) | Video classification methods, device, equipment and medium based on artificial intelligence | |
CN111027419B (en) | Method, device, equipment and medium for detecting video irrelevant content | |
CN110531849A (en) | Intelligent teaching system based on 5G communication and capable of enhancing reality | |
CN113963304B (en) | Cross-modal video time sequence action positioning method and system based on time sequence-space diagram | |
CN113393544B (en) | Image processing method, device, equipment and medium | |
CN113723530A (en) | Intelligent psychological assessment system based on video analysis and electronic psychological sand table | |
CN116309992A (en) | Intelligent meta-universe live person generation method, equipment and storage medium | |
CN111405314B (en) | Information processing method, device, equipment and storage medium | |
Cui et al. | Deep learning based advanced spatio-temporal extraction model in medical sports rehabilitation for motion analysis and data processing | |
Liu et al. | Trampoline motion decomposition method based on deep learning image recognition | |
He | Athlete human behavior recognition based on continuous image deep learning and sensors | |
CN115546491B (en) | Fall alarm method, system, electronic equipment and storage medium | |
CN111783587A (en) | Interaction method, device and storage medium | |
CN108596068A (en) | A kind of method and apparatus of action recognition | |
Zhao | Research on athlete behavior recognition technology in sports teaching video based on deep neural network | |
CN111760276B (en) | Game behavior control method, device, terminal, server and storage medium | |
CN113824989A (en) | Video processing method and device and computer readable storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190927 |