Emotion recognition analysis method for expression data based on artificial intelligence

Abstract

The invention discloses an emotion recognition analysis method for expression data based on artificial intelligence, and relates to the technical field of image recognition, the method comprises the steps: collecting the facial contour of a target person, constructing a facial model, selecting a plurality of traction points corresponding to the facial model, and carrying out the combination according to the corresponding emotions of the traction points, and obtaining a combined point cloud of a plurality of corresponding emotions; establishing a pull line relationship between the combined point clouds of the plurality of corresponding emotions and the face model; collecting face emotional expression video image data of the target person in a preset time period, and fitting the face emotional expression video image data with the face model to obtain a fitting model; and based on the face model in the fitting model, obtaining a combined point cloud of a corresponding emotion through a pull line relationship. According to the method, the emotion recognition data can be embodied and cleared, the emotion types which do not correspond to the target person can be deleted and selected in the dynamic analysis process of the face emotion expression video image data of the target person, and the accuracy of emotion type recognition can be ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to an emotion recognition and analysis method of expression data based on artificial intelligence. Background Art

[0002] With the continuous advancement of technology and the growing demand for applications, the application scenarios of emotion recognition and analysis methods based on facial expression data are also expanding. In addition to human-computer interaction, social media analysis, and mental health, this technology can also be applied to entertainment, security monitoring, and other fields. At present, due to the complexity and diversity of human emotions, as well as the ambiguity and uncertainty of facial expression data, which will affect the inaccuracy of character emotion recognition, it is difficult to clearly identify facial expression data. Summary of the invention

[0003] The purpose of the present invention is to provide an emotion recognition and analysis method for facial expression data based on artificial intelligence to solve the shortcomings of the background technology.

[0004] In order to achieve the above object, the present invention provides the following technical solution: an emotion recognition and analysis method of expression data based on artificial intelligence, comprising the following steps:

[0005] The facial contour of the target person is collected and a facial model is constructed. Multiple driving points are selected according to the facial model, and the driving points are combined according to the corresponding emotions to obtain a plurality of combined point clouds corresponding to the emotions.

[0006] Establishing a connection relationship between multiple combined point clouds corresponding to emotions and the facial model;

[0007] Collecting video image data of facial emotion expressions of a target person within a preset time period, and fitting the video image data of facial emotion expressions to a facial model to obtain a fitting model;

[0008] Based on the facial model in the fitting model, the combined point cloud corresponding to the emotion is obtained through the matching relationship to determine the emotion type of the target person.

[0009] In a preferred embodiment, the steps of collecting the facial contour of the target person and constructing a facial model, selecting a plurality of driving points corresponding to the facial model, and combining the driving points according to the corresponding emotions to obtain a plurality of combined point clouds corresponding to the emotions include:

[0010] Acquire a facial contour image of a target person from multiple directions, and construct a facial model based on the facial contour image of the target person;

[0011] Selecting a plurality of affecting points on the facial model based on facial muscle composition information, wherein the affecting points are feature points corresponding to facial emotion muscles;

[0012] Multiple driving points corresponding to the emotion types are combined to obtain multiple combined point clouds.

[0013] In a preferred embodiment, the step of establishing a connection relationship between a plurality of combined point clouds corresponding to emotions and the facial model comprises:

[0014] The facial model is copied to obtain a plurality of emotional facial models, wherein the number of the emotional facial models is the same as the number of the combined point clouds, and the combined point clouds are bound to the emotional facial models in a one-to-one correspondence;

[0015] Based on the emotional face model, motion trajectory information of a plurality of corresponding emotional combined point clouds is respectively established, wherein the motion trajectory information includes the movement trajectories of a plurality of driving points in the combined point cloud according to the emotion type and the positional relationship of the movement trajectories between the plurality of driving points in the combined point cloud;

[0016] On the emotional face model, multiple combined point clouds are respectively constructed with corresponding motion trajectory information to obtain multiple facial emotion point cloud sets;

[0017] Establish a connection relationship between multiple facial emotion point cloud sets and facial models.

[0018] In a preferred embodiment, the step of establishing motion trajectory information of a plurality of corresponding emotion combination point clouds based on the emotion facial model comprises:

[0019] Determine the affected position information of the multiple affected points of the combined point cloud according to the emotion type on the emotional face model, wherein the affected position information includes the affected final positions of the multiple affected points and the positional relationship between the multiple affected points when the affected final positions are reached;

[0020] Determine the paths from the original positions of the multiple affected points in the combined point cloud to the final affected positions as the movement trajectories of the multiple affected points;

[0021] The moving trajectories of the multiple affecting points in the combined point cloud according to the emotion type and the positional relationship of the moving trajectories between the multiple affecting points in the combined point cloud are used as the motion trajectory information.

[0022] In a preferred embodiment, the step of establishing a connection relationship between a plurality of facial emotion point cloud sets and a facial model comprises:

[0023] Assign corresponding ports to multiple moving points in the facial model respectively;

[0024] Corresponding sub-ports are respectively configured for multiple affecting points in multiple facial emotion point cloud sets, and a position attachment relationship between the sub-ports and the moving trajectories of the corresponding affecting points is established;

[0025] The sub-ports with established attachment relationships are linearly connected to the ports of the corresponding driving points in the facial model to obtain the connection relationship between the multiple facial emotion point clouds and the facial model.

[0026] In a preferred embodiment, the step of collecting the facial emotion expression video image data of the target person within a preset time period, fitting the facial emotion expression video image data with the facial model, and obtaining the fitted model includes:

[0027] Collecting video image data of facial emotion expressions of a target person within a preset time period, wherein the video image data of facial emotion expressions includes a facial contour in the video image of facial emotion expressions and a moving point in the facial contour;

[0028] The facial contour in the facial emotion expression video image data is overlapped with the facial model, and the moving points in the facial contour in the facial emotion expression video image data are adjusted to fit the corresponding moving points in the facial model to obtain a fitted model.

[0029] In a preferred embodiment, the step of obtaining a combined point cloud corresponding to an emotion through a tie-line relationship based on the facial model in the fitting model to determine the emotion type of the target person includes:

[0030] Binding the moving points in the facial contour in the fitted facial emotion expression video image data with the corresponding moving points in the facial model;

[0031] The moving point in the facial model moves along with the moving point in the facial contour in the bound human facial emotion expression video image data, and the moving speed of the moving point in the facial contour in the human facial emotion expression video image data is obtained;

[0032] The multiple driving points in the multiple facial emotion point cloud sets are started to move dynamically according to the moving speed, and the multiple driving points in the multiple facial emotion point cloud sets that move dynamically are connected with the driving points in the facial model based on the line relationship to obtain the recognition indexes of the facial model corresponding to the multiple facial emotion point cloud sets respectively;

[0033] Determine the combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset conditions, and obtain the emotion type corresponding to the target person according to the combined point cloud.

[0034] In a preferred embodiment, the step of obtaining the recognition indexes of the facial model corresponding to the plurality of facial emotion point clouds respectively comprises:

[0035] Obtaining the lengths and angles of the lines connecting the multiple driving points in the multiple facial emotion point cloud sets that are dynamically moving and the driving points in the facial model;

[0036] The recognition indexes of the facial models corresponding to the multiple facial emotion point cloud sets are obtained respectively. The calculation formula of the recognition index includes:

[0037] Among them, θ is the recognition index, LY n LD is the original length of the line connecting the nth driving point in a single facial emotion point cloud and the corresponding driving point in the facial model. n is the current length of the line connecting the nth driving point in a single facial emotion point cloud set and the corresponding driving point in the facial model, μY n is the original angle between the nth driving point in a single facial emotion point cloud and the driving point line in the corresponding facial model, μD n is the current angle between the nth affecting point in a single facial emotion point cloud set and the affecting point line in the corresponding facial model, n is the nth affecting point among the k affecting points in a single facial emotion point cloud set, wherein α and β are constants greater than zero.

[0038] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0039] The present invention calculates the recognition index in real time during the process of connecting multiple driving points in multiple dynamically moving facial emotion point cloud sets with driving points in the facial model by lines, and collecting the lengths and angles of the line connections between multiple driving points in multiple facial emotion point cloud sets and driving points in the facial model. It can concretize and clarify the emotion recognition data, and then determine the combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset conditions, and obtain the emotion type corresponding to the target person according to the combined point cloud. In the process of dynamically analyzing the facial emotion expression video image data of the target person, the emotion type that does not correspond to the target person can be deleted, and efficient recognition can be achieved, and the accuracy of emotion type recognition can be guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0041] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Example 1, please refer to Figure 1 As shown, the emotion recognition and analysis method of expression data based on artificial intelligence described in this embodiment includes the following steps:

[0044] S1. Collect the facial contour of the target person and build a facial model. Select multiple moving points corresponding to the facial model, and combine the moving points according to the corresponding emotions to obtain a combination point cloud corresponding to multiple emotions.

[0045] S2, establishing a connection relationship between a plurality of combined point clouds corresponding to emotions and a facial model;

[0046] S3, collecting video image data of facial emotion expressions of the target person within a preset time period, and fitting the video image data of facial emotion expressions with the facial model to obtain a fitting model;

[0047] S4, obtaining a combined point cloud corresponding to the emotion through the tie-line relationship based on the facial model in the fitting model, and determining the emotion type of the target person;

[0048] As described in the above steps S1-S4, with the continuous advancement of technology and the continuous growth of application needs, the application scenarios of the emotion recognition and analysis method through expression data are also constantly expanding. In addition to the fields of human-computer interaction, social media analysis and mental health, this technology can also be applied to entertainment, security monitoring and other fields. At present, due to the complexity and diversity of human emotions, as well as the ambiguity and uncertainty of expression data, the inaccuracy of character emotion recognition will be affected, and it is difficult to clearly identify the expression data. In this application, multiple moving points in multiple facial emotion point cloud sets that are dynamically moving are respectively connected with the moving points in the facial model by lines, and the lengths and angles of the lines connecting multiple moving points in multiple facial emotion point cloud sets and the moving points in the facial model are collected. The recognition index is calculated in real time, which can concretize and clarify the emotion recognition data, and then determine the combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset conditions, and obtain the emotion type corresponding to the target person according to the combined point cloud. In the process of dynamically analyzing the facial emotion expression video image data of the target person, the emotion type that does not correspond to the target person can be deleted, which can be efficiently recognized and the accuracy of emotion type recognition can be guaranteed.

[0049] In one embodiment, the step S1 of collecting the facial contour of the target person and constructing a facial model, selecting a plurality of driving points corresponding to the facial model, and combining the driving points according to the corresponding emotions to obtain a plurality of combined point clouds corresponding to the emotions comprises:

[0050] S11, acquiring a facial contour image of a target person from multiple directions, and constructing a facial model based on the facial contour image of the target person;

[0051] S12, selecting a plurality of affecting points on the facial model based on the facial muscle composition information, wherein the affecting points are feature points corresponding to facial emotion muscles;

[0052] S13, combining the multiple affect points corresponding to the emotion types to obtain multiple combined point clouds;

[0053] As described in the above steps S11-S13, before the target person is subjected to emotion recognition, the facial contour information of the target person needs to be collected. The collection method is as follows: using a high-quality camera or camera to collect facial images to ensure that the images are clear, blur-free and noise-free, collecting facial contour images of the target person from multiple angles, converting color images into grayscale images, using filters to remove noise in the images, improving image quality, and making facial contours clearer through contrast enhancement, sharpening and other technical means, using computer vision technology, such as the shape predictor in the dlib library, to detect facial key points, such as eyes, nose, mouth, etc., matching the detected feature points with the feature points in the predefined facial model to provide a basis for subsequent three-dimensional reconstruction, using a set of two-dimensional facial images and corresponding feature point information, and constructing a three-dimensional facial model through a three-dimensional reconstruction algorithm (such as 3DMM), and then in order to understand the emotions of the target person, it is necessary to analyze the facial muscles that affect the emotions to obtain multiple affect points, The driving points are feature points corresponding to facial muscles, wherein the feature points corresponding to facial muscles are points where muscles drive the facial contour to move when emotions appear on the face. For example, when smiling happily, the position of the highest point of the muscles around the cheekbones and the muscles around the mouth change. Here, the corresponding muscle point mark is used as a feature point, and then as a driving point. This operation depends on the understanding of the composition information of the human facial structure. Therefore, multiple driving points are selected on the facial model based on the facial muscle composition information, and then the multiple driving points are combined according to the emotion type to obtain multiple combined point clouds. For example, the muscles involved in a happy smile mainly include zygomatic muscle, buccinator muscle, orbicularis oris muscle, zygomatic major muscle, etc. Therefore, it is necessary to combine the driving points corresponding to the zygomatic muscle, buccinator muscle, orbicularis oris muscle, zygomatic major muscle, etc. corresponding to the emotion type to obtain a combined point cloud. Multiple emotion types have corresponding muscle driving points, so they can be combined separately according to the emotion types, which can facilitate the preparation of emotion recognition for the target person.

[0054] In one embodiment, the step S2 of establishing a connection relationship between a plurality of combined point clouds corresponding to emotions and a facial model comprises:

[0055] S21, duplicating the facial model to obtain a plurality of emotional facial models, wherein the number of the emotional facial models is the same as the number of the combined point clouds, and binding the combined point clouds to the emotional facial models in a one-to-one correspondence;

[0056] S22, establishing motion trajectory information of a plurality of corresponding emotion combined point clouds respectively based on the emotion facial model, wherein the motion trajectory information includes the movement trajectories of a plurality of driving points in the combined point cloud according to the emotion type and the positional relationship of the movement trajectories between the plurality of driving points in the combined point cloud;

[0057] S23, constructing information of the multiple combined point clouds and the corresponding motion trajectory information on the emotional facial model to obtain multiple facial emotion point cloud sets;

[0058] S24. Establishing a connection relationship between multiple facial emotion point clouds and the facial model.

[0059] As described in the above steps S21-S23, since each emotion type needs to be displayed separately, the facial model needs to be copied to obtain multiple emotion facial models, wherein the number of emotion facial models is the same as the number of combined point clouds, and the combined point clouds are bound to the emotion facial models in a one-to-one correspondence. The combined point clouds of the corresponding emotion types can be displayed through the emotion facial models, which can better facilitate the recognition and confirmation of subsequent emotions, and establish motion trajectory information of multiple corresponding emotion combined point clouds. The motion trajectory information here is the movement trajectory of the driving points of the corresponding emotion types in the combined point cloud and the positional relationship between the movement trajectories of multiple driving points in the combined point cloud, so that the multiple combined point clouds under the corresponding emotion types can be obtained. The motion trajectory of the moving point and the positional relationship between the moving trajectories of multiple moving points in the combined point cloud can help us understand the changes in the facial muscles of the target person under different emotional types, which is convenient for the subsequent recognition of the target person's emotions. After that, the multiple combined point clouds are respectively constructed with the corresponding motion trajectory information. The construction here is to construct a three-dimensional trajectory of the combined point cloud on the facial model according to the motion trajectory information to obtain multiple facial emotion point cloud sets; then, the connection relationship between the multiple facial emotion point cloud sets and the facial model is established, which can facilitate the corresponding relationship between the subsequently collected facial emotion expression video image data and the multiple facial emotion point cloud sets, and can better perform subsequent expression matching and recognition, thereby improving the efficiency and accuracy of expression recognition;

[0060] In one embodiment, the step S22 of establishing motion trajectory information of a plurality of corresponding emotion combination point clouds based on the emotion facial model comprises:

[0061] S221, determining the affected position information of the multiple affected points of the combined point cloud according to the emotion type on the emotional facial model, wherein the affected position information includes the affected final positions of the multiple affected points and the positional relationship between the multiple affected points when the affected final positions are reached;

[0062] S222, determining a path from the original position of the multiple affecting points in the combined point cloud to the final affecting position as a moving trajectory of the multiple affecting points;

[0063] S223, taking the moving trajectories of the multiple affecting points in the combined point cloud according to the emotion type and the positional relationship of the moving trajectories between the multiple affecting points in the combined point cloud as the motion trajectory information;

[0064] As described in the above steps S221-S223, when the target person has an expression, the position of the attraction point will change. Therefore, based on the combined point cloud, the final attraction position of the multiple attraction points in the combined point cloud is determined according to the emotion type, and the positional relationship between the multiple attraction points when the final attraction position is determined. After the final attraction position of the attraction point in the combined point cloud according to the emotion type is determined, it can be displayed on the corresponding emotional facial model. Then, according to the attraction position information and the original positions of the multiple attraction points on the corresponding emotional facial model, the paths of moving the original positions of the multiple attraction points to the final attraction positions are determined, and the paths are respectively moved. As the moving trajectories of multiple driving points, the moving trajectories of the driving points in the combined point cloud according to the emotion type and the positional relationship of the moving trajectories between the multiple driving points in the combined point cloud are then used as motion trajectory information, so as to facilitate the subsequent dynamic construction on the corresponding emotion facial model through the motion trajectory information, and obtain the multiple driving points that dynamically move at the original position according to the motion trajectory information, and then obtain the subsequent facial emotion point cloud set. The multiple combined point clouds can construct multiple facial emotion point cloud sets, and multiple emotion types can be constructed separately, so as to facilitate the subsequent rapid emotion recognition of the actual target person's emotion generation relationship;

[0065] In one embodiment, the step S24 of establishing a tie relationship between a plurality of facial emotion point cloud sets and a facial model comprises:

[0066] S241, configuring corresponding ports for multiple moving points in the facial model respectively;

[0067] S242, respectively configuring corresponding sub-ports corresponding to the multiple affecting points in the multiple facial emotion point cloud sets, and establishing a position attachment relationship between the sub-ports and the moving trajectories of the corresponding affecting points;

[0068] S243, linearly connecting the sub-ports with established attachment relationships to the ports of the corresponding driving points in the facial model, to obtain the connection relationship between the multiple facial emotion point clouds and the facial model;

[0069] As described in the above steps S241-S243, the facial model is a model used to display the actual facial condition of the target person. Since the facial model is constructed according to the facial contour of the target person, it can correspond to the facial action points subsequently collected by the target person. Here, corresponding ports are configured for the multiple action points in the facial model, and the ports are used to connect and act with the corresponding action points in the subsequent facial emotion point cloud set. Then, corresponding sub-ports are configured for the multiple action points in the corresponding facial emotion point cloud sets. Since the facial emotion point cloud set is a dynamic model for displaying the corresponding emotion type, the sub-ports are configured for the action points in the corresponding facial emotion point cloud set. Therefore, the sub-ports are also dynamic. The sub-ports here are dynamic. The embodiment is to attach the sub-port to the moving trajectory, realize the movement of the sub-port according to the movement of the corresponding driving point, and linearly connect the sub-port with the corresponding port on the facial model. The linear connection here represents the connection length between the sub-port and the port. The port here is also moved at the corresponding driving point in the subsequent corresponding facial model. Before the actual target person's facial video image data is collected, the port here is fixed in place following the corresponding driving point in the facial model. Since there are repeated driving points in multiple combined point clouds, the driving points in the facial model are connected to multiple sub-ports. Here, the port can be connected to multiple sub-ports at the same time, which can more efficiently and accurately identify the facial emotions of the target person.

[0070] In one embodiment, the step S3 of collecting the facial emotion expression video image data of the target person within a preset time period, fitting the facial emotion expression video image data with the facial model, and obtaining the fitted model includes:

[0071] S31, collecting facial emotion expression video image data of the target person within a preset time period, wherein the facial emotion expression video image data includes a facial contour in the facial emotion expression video image and a moving point in the facial contour;

[0072] S32, overlapping the facial contour in the facial emotion expression video image data with the facial model, adjusting the moving points in the facial contour in the facial emotion expression video image data to fit the corresponding moving points in the facial model, and obtaining a fitted model;

[0073] As described in the above steps S31 and S32, the facial emotion expression video image data within a preset time period is collected. The preset time period here is not a fixed time period, but a change node when the target person produces different emotional expressions. For example, the current person is smiling happily. When the happy smile ends and reaches a peaceful emotional expression, it is used as the collection starting point. After the peaceful emotional expression ends and reaches the happy smile, the collection of the facial emotion expression video image data within the current preset time period is completed. Then, continue with the start of the happy smile emotional expression as the collection starting point. After the happy smile emotional expression ends, the collection of the facial emotion expression video image data within the next preset time period is completed. The facial emotion expression video image data of the target person is collected every moment. The point where different emotional expressions end is used as a collection node, and the facial emotional expression video image data includes the facial contour in the facial emotional expression video image and the driving points in the facial contour. Because there is mostly a calm process of facial expression during the expression conversion, such as from a smiling expression to a calm expression, and then a smiling expression again, the starting point of the facial emotional expression video image data collected within the preset time period is usually calm, and there are some errors that require adjustment of the driving points in the facial contour in the facial emotional expression video image data until the driving points in the facial contour in the facial emotional expression video image data fit (overlap) with the driving points in the corresponding facial model, thereby completing the recognition preparation of the target person's emotions in the facial emotional expression video image data within the preset time period.

[0074] In one embodiment, the step S4 of obtaining a combined point cloud corresponding to an emotion through a tie-line relationship based on the facial model in the fitting model to determine the emotion type of the target person includes:

[0075] S41, binding the moving points in the facial contour in the fitted facial emotion expression video image data with the corresponding moving points in the facial model;

[0076] S42, the moving point in the facial model moves along with the moving point in the facial contour in the bound facial emotion expression video image data, and the moving speed of the moving point in the facial contour in the facial emotion expression video image data is obtained;

[0077] S43, starting multiple driving points in multiple facial emotion point cloud sets to move dynamically according to the moving speed, and connecting multiple driving points in multiple facial emotion point cloud sets that move dynamically with driving points in the facial model based on the line relationship, to obtain recognition indexes of the facial model corresponding to the multiple facial emotion point cloud sets respectively;

[0078] S44, determining a combined point cloud of a facial emotion point cloud set corresponding to a recognition index that meets a preset condition, and obtaining an emotion type corresponding to the target person according to the combined point cloud;

[0079] As described in the above steps S41-S44, when the driving points in the facial contour in the facial emotion expression video image data are fitted with the corresponding driving points in the facial model, they are bound so that the driving points in the facial model can move along with the driving points in the facial contour in the corresponding facial emotion expression video image data, and the moving speed of the driving points in the facial contour in the facial emotion expression video image data is notified, and the multiple driving points in the multiple facial emotion point cloud sets are started to move dynamically according to the moving speed. Then, at the same moving speed, if the recognition index obtained by the length and angle of the line connection between the multiple driving points in the multiple facial emotion point cloud sets that are dynamically moved and the driving points in the facial model is larger, it means that the corresponding facial emotion point cloud set does not correspond to the actual emotion type of the target person, and the multiple driving points in the multiple facial emotion point cloud sets that are dynamically moved are connected to the driving points in the facial model based on the string relationship. The line connection here is similar to that of a puppet, and the movement of the string is the same as that of the puppet. The line here represents the multiple facial emotion point cloud sets. The multiple driving points are respectively connected with the lines between the driving points in the facial model. If they move synchronously, the length and angle of the line connection will not change or will change slightly within the range. Here, through the preset condition restriction, within the preset time period, the combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset condition is determined, and the emotion type corresponding to the target person is obtained according to the combined point cloud; the multiple driving points in the multiple facial emotion point cloud sets that are dynamically moved are respectively connected with the driving points in the facial model, and the lengths and angles of the line connections between the multiple driving points in the multiple facial emotion point cloud sets and the driving points in the facial model are collected, and the recognition index is calculated in real time, which can concretize and clarify the emotion recognition data, and then the combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset condition is determined, and the emotion type corresponding to the target person is obtained according to the combined point cloud. In the process of dynamically analyzing the facial emotion expression video image data of the target person, the emotion type that does not correspond to the target person can be deleted, which can be efficiently recognized and the accuracy of emotion type recognition can be guaranteed.

[0080] In one embodiment, the step S43 of obtaining the recognition indexes of the facial models corresponding to the plurality of facial emotion point cloud sets respectively comprises:

[0081] S431, obtaining the lengths and angles of the lines connecting the multiple moving points in the multiple facial emotion point cloud sets that are dynamically moving and the moving points in the facial model;

[0082] S432, respectively obtain recognition indexes of the facial models corresponding to the plurality of facial emotion point cloud sets, wherein the calculation formula of the recognition index includes:

[0083] Among them, θ is the recognition index, LY n LD is the original length of the line connecting the nth driving point in a single facial emotion point cloud and the corresponding driving point in the facial model. n is the current length of the line connecting the nth driving point in a single facial emotion point cloud set and the corresponding driving point in the facial model, μY n is the original angle between the nth driving point in a single facial emotion point cloud and the driving point line in the corresponding facial model, μD n is the current angle between the nth affecting point in a single facial emotion point cloud set and the affecting point line in the corresponding facial model, n is the nth affecting point among the k affecting points in a single facial emotion point cloud set, wherein α and β are constants greater than zero. It should be noted that the larger the θ is, the lower the matching degree is. The combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset conditions is selected, and the emotion type corresponding to the combined point cloud is taken as the emotion type of the target person.

[0084] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An emotion recognition and analysis method for facial expression data based on artificial intelligence, characterized in that: The following steps are involved: The facial contour of the target person is collected and a facial model is constructed. Multiple moving points are selected according to the facial model, and the moving points are combined according to the corresponding emotions to obtain a plurality of combined point clouds corresponding to the emotions. Establishing a connection relationship between multiple combined point clouds corresponding to emotions and the facial model; Collecting video image data of facial emotion expressions of a target person within a preset time period, and fitting the video image data of facial emotion expressions to a facial model to obtain a fitting model; Based on the facial model in the fitting model, the combined point cloud corresponding to the emotion is obtained through the matching relationship to determine the emotion type of the target person.

2. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 1, characterized in that: The steps of collecting the facial contour of the target person and constructing a facial model, selecting a plurality of driving points corresponding to the facial model, and combining the driving points according to the corresponding emotions to obtain a plurality of combined point clouds corresponding to the emotions include: Acquire a facial contour image of a target person from multiple directions, and construct a facial model based on the facial contour image of the target person; Selecting a plurality of affecting points on the facial model based on facial muscle composition information, wherein the affecting points are feature points corresponding to facial emotion muscles; Multiple driving points corresponding to the emotion types are combined to obtain multiple combined point clouds.

3. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 1, characterized in that: The step of establishing a connection relationship between a plurality of combined point clouds corresponding to emotions and the facial model comprises: The facial model is copied to obtain a plurality of emotional facial models, wherein the number of the emotional facial models is the same as the number of the combined point clouds, and the combined point clouds are bound to the emotional facial models in a one-to-one correspondence; Based on the emotional face model, motion trajectory information of a plurality of corresponding emotional combined point clouds is respectively established, wherein the motion trajectory information includes the movement trajectories of a plurality of driving points in the combined point cloud according to the emotion type and the positional relationship of the movement trajectories between the plurality of driving points in the combined point cloud; On the emotional face model, multiple combined point clouds are respectively constructed with corresponding motion trajectory information to obtain multiple facial emotion point cloud sets; Establish a connection relationship between multiple facial emotion point cloud sets and facial models.

4. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 3, characterized in that: The step of establishing motion trajectory information of a plurality of corresponding emotion combination point clouds based on the emotion facial model comprises: Determine the affected position information of the multiple affected points of the combined point cloud according to the emotion type on the emotional face model, wherein the affected position information includes the affected final positions of the multiple affected points and the positional relationship between the multiple affected points when the affected final positions are reached; Determine the paths from the original positions of the multiple affected points in the combined point cloud to the final affected positions as the movement trajectories of the multiple affected points; The moving trajectories of the multiple affecting points in the combined point cloud according to the emotion type and the positional relationship of the moving trajectories between the multiple affecting points in the combined point cloud are used as the motion trajectory information.

5. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 3, characterized in that: The step of establishing a connection relationship between a plurality of facial emotion point cloud sets and a facial model comprises: Assign corresponding ports to multiple moving points in the facial model respectively; Corresponding sub-ports are respectively configured for multiple affecting points in multiple facial emotion point cloud sets, and a position attachment relationship between the sub-ports and the moving trajectories of the corresponding affecting points is established; The sub-ports with established attachment relationships are linearly connected to the ports of the corresponding driving points in the facial model to obtain the connection relationship between the multiple facial emotion point clouds and the facial model.

6. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 1, characterized in that: The step of collecting the facial expression video image data of the target person within a preset time period, fitting the facial expression video image data with the facial model, and obtaining the fitting model includes: Collecting video image data of facial emotion expressions of a target person within a preset time period, wherein the video image data of facial emotion expressions includes a facial contour in the video image of facial emotion expressions and a moving point in the facial contour; The facial contour in the facial emotion expression video image data is overlapped with the facial model, and the moving points in the facial contour in the facial emotion expression video image data are adjusted to fit the corresponding moving points in the facial model to obtain a fitted model.

7. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 1, characterized in that: The step of obtaining a combined point cloud corresponding to the emotion through a tie-line relationship based on the facial model in the fitting model and determining the emotion type of the target person comprises: Binding the moving points in the facial contour in the fitted facial emotion expression video image data with the corresponding moving points in the facial model; The moving point in the facial model moves along with the moving point in the facial contour in the bound human facial emotion expression video image data, and the moving speed of the moving point in the facial contour in the human facial emotion expression video image data is obtained; The multiple driving points in the multiple facial emotion point cloud sets are started to move dynamically according to the moving speed, and the multiple driving points in the multiple facial emotion point cloud sets that move dynamically are connected with the driving points in the facial model based on the line relationship to obtain the recognition indexes of the facial model corresponding to the multiple facial emotion point cloud sets respectively; Determine the combined point cloud of the facial emotion point cloud set corresponding to the recognition index that meets the preset conditions, and obtain the emotion type corresponding to the target person according to the combined point cloud.

8. The emotion recognition and analysis method of facial expression data based on artificial intelligence according to claim 1, characterized in that: The step of obtaining the recognition indexes of the facial model corresponding to the plurality of facial emotion point cloud sets respectively comprises: Obtaining the lengths and angles of the lines connecting the multiple driving points in the multiple facial emotion point cloud sets that are dynamically moving and the driving points in the facial model; The recognition indexes of the facial models corresponding to the multiple facial emotion point cloud sets are obtained respectively. The calculation formula of the recognition index includes: Among them, θ is the recognition index, LY n LD is the original length of the line connecting the nth driving point in a single facial emotion point cloud and the corresponding driving point in the facial model. n is the current length of the line connecting the nth driving point in a single facial emotion point cloud set and the corresponding driving point in the facial model, μY n is the original angle between the nth driving point in a single facial emotion point cloud and the driving point line in the corresponding facial model, μD n is the current angle between the nth affecting point in a single facial emotion point cloud set and the affecting point line in the corresponding facial model, n is the nth affecting point among the k affecting points in a single facial emotion point cloud set, wherein α and β are constants greater than zero.

Images