Digital human-based abnormal expression recognition method and device

Abstract

The present disclosure provides an abnormal expression recognition method and device based on a digital person, wherein the method comprises: generating first facial expression parameters corresponding to each frame expression in expression information according to input facial driving information; processing the first facial expression parameters corresponding to each frame expression to generate a corresponding first expression image; inputting the first expression image into a pre-trained evaluation model to obtain second facial expression parameters; processing the second facial expression parameters to generate a corresponding second expression image; calculating the similarity of the first expression image and the second expression image corresponding to each frame expression, and marking the frame expression as an abnormal expression if it is determined that the similarity is less than a threshold value, so that the abnormal expression can be automatically identified, the burden of the test personnel is relieved, and the processing efficiency is improved.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to a method and apparatus for recognizing abnormal facial expressions based on digital humans. Background Technology

[0002] Digital humans are virtual characters with digital appearances that exist on display devices and possess human features and mannerisms.

[0003] Taking digital human facial expressions as an example, a digital human-driven algorithm is needed to generate animated facial expressions based on input test data, thereby giving the digital human various expressions. In existing technologies, evaluating the effectiveness of digital human-driven algorithms relies on the subjective evaluation of testers. Testers need to carefully examine every frame of the digital human's animated facial expressions to check for any unusual or strange expressions. This process is extremely time-consuming, labor-intensive, and inefficient. Summary of the Invention

[0004] This disclosure provides a method and apparatus for abnormal facial expression recognition based on digital humans, in order to solve the technical defects of the prior art, which are time-consuming, labor-intensive, and inefficient in abnormal image screening.

[0005] In a first aspect, embodiments of this disclosure provide a method for recognizing abnormal facial expressions based on digital humans, including:

[0006] Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of expression in the expression information;

[0007] The first facial expression parameters corresponding to each frame of expression are processed to generate the corresponding first expression image.

[0008] The first facial expression image is input into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images;

[0009] The second facial expression parameters are processed to generate the corresponding second expression image;

[0010] Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression.

[0011] According to the abnormal expression recognition method based on digital humans provided in this disclosure, the method generates first facial expression parameters corresponding to each frame of expression information based on the input facial driving information, including:

[0012] The facial driving information is input into a pre-trained facial driving model to generate first facial expression parameters corresponding to each frame of expression information; wherein, the facial driving information includes at least one of image, video, and text;

[0013] The facial driving model is trained using sample driving information and sample expression parameters.

[0014] According to the abnormal facial expression recognition method based on digital humans provided in this disclosure, the method processes the first facial expression parameters corresponding to each frame of expression to generate a corresponding first expression image, including:

[0015] The first facial expression parameters corresponding to each frame of expression are input into the rendering model to generate the first three-dimensional digital human image.

[0016] The first three-dimensional digital human image is projected onto a two-dimensional plane to generate the first facial expression image.

[0017] According to the abnormal facial expression recognition method based on digital humans provided in this disclosure, the method processes second facial expression parameters to generate a corresponding second expression image, including:

[0018] The second facial expression parameters corresponding to each frame of expression are input into the rendering model to generate a second three-dimensional digital human image.

[0019] The second three-dimensional digital human image is projected onto a two-dimensional plane to generate the second facial expression image.

[0020] According to the abnormal expression recognition method based on digital humans provided in this disclosure, after determining that the expression in the frame is an abnormal expression, the method further includes:

[0021] The expression in the frame is corrected to obtain the corrected expression information, and the digital human's expression animation is generated based on the corrected expression information.

[0022] Secondly, embodiments of this disclosure provide an abnormal facial expression recognition device based on a digital human, comprising:

[0023] The expression parameter generation module is used to generate the first facial expression parameters corresponding to each frame of expression information based on the input facial driving information.

[0024] The first image processing module is used to process the first facial expression parameters corresponding to each frame of expression to generate the corresponding first expression image.

[0025] An evaluation module is used to input the first facial expression image into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images;

[0026] The second image processing module is used to process the second facial expression parameters and generate the corresponding second expression image.

[0027] The labeling module is used to calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression, and to label the expression in the frame as an abnormal expression if the similarity is less than a threshold.

[0028] According to the abnormal facial expression recognition device based on digital human provided in this disclosure, the first image processing module is specifically used for:

[0029] The first facial expression parameters corresponding to each frame of expression are input into the rendering model to generate the first three-dimensional digital human image.

[0030] The first three-dimensional digital human image is projected onto a two-dimensional plane to generate the first facial expression image.

[0031] According to the abnormal facial expression recognition device based on digital human provided in this disclosure, the second image processing module is specifically used for:

[0032] The second facial expression parameters corresponding to each frame of expression are input into the rendering model to generate a second three-dimensional digital human image.

[0033] The second three-dimensional digital human image is projected onto a two-dimensional plane to generate the second facial expression image.

[0034] Thirdly, this disclosure provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the abnormal facial expression recognition method based on digital human as described in any of the preceding claims.

[0035] Fourthly, this disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the abnormal facial expression recognition method based on digital humans as described in any of the preceding claims.

[0036] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the abnormal facial expression recognition method based on digital humans as described in any of the preceding claims.

[0037] The present disclosure provides a method and apparatus for abnormal facial expression recognition based on digital humans. Based on input facial driving information, it generates first facial expression parameters corresponding to each frame of facial expression information; processes these first facial expression parameters to generate a corresponding first facial expression image; inputs the first facial expression image into a pre-trained evaluation model to obtain second facial expression parameters; processes these second facial expression parameters to generate a corresponding second facial expression image; calculates the similarity between the first and second facial expression images corresponding to each frame of facial expression; and marks the frame of facial expression as an abnormal expression if the similarity is less than a threshold. This allows for automatic identification of abnormal facial expressions, reducing the burden on testers and improving processing efficiency. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is one of the flowcharts illustrating the abnormal facial expression recognition method based on digital humans provided in this embodiment of the disclosure;

[0040] Figure 2 This is the second flowchart illustrating the abnormal facial expression recognition method based on digital humans provided in this embodiment of the disclosure;

[0041] Figure 3 This is a schematic diagram of a first facial expression image generated according to the abnormal facial expression recognition method provided in the embodiments of this disclosure;

[0042] Figure 4 This is a schematic diagram of a second facial expression image generated according to the abnormal facial expression recognition method provided in the embodiments of this disclosure;

[0043] Figure 5 This is a schematic diagram of the structure of the abnormal facial expression recognition device based on digital human provided in the embodiments of this disclosure;

[0044] Figure 6 This is a schematic diagram of the structure of the electronic device provided in this disclosure. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this disclosure clearer, the technical solutions of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0046] This disclosure includes an embodiment of a method for recognizing abnormal facial expressions based on digital humans, see [link to relevant documentation]. Figure 1 This includes the following steps 101 to 105:

[0047] Step 101: Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of the expression information.

[0048] Specifically, facial driving information includes at least one of text, voice, image, and video. Through facial driving information, digital humans can change their facial expressions according to the corresponding text, voice, or other content to achieve broadcasting. For example, in one use case, a digital human can read aloud an article based on the input.

[0049] Facial driving information includes timestamp information. Based on the timestamp information, first facial expression parameters corresponding to multiple frames of expressions can be generated respectively.

[0050] The first facial expression parameters can include various parameters, such as eye movement parameters, eyelid movement parameters, and lip movement parameters. These parameters enable the digital human to display a variety of facial expressions, such as joy, anger, sorrow, and happiness.

[0051] Step 102: Process the first facial expression parameters corresponding to each frame of expression to generate the corresponding first expression image.

[0052] There are several ways to process facial expression parameters, such as through rendering models in rendering software, like Maya and Wrap3.

[0053] Specifically, in one implementation, step 102 includes: inputting the first facial expression parameters corresponding to each frame of expression into the rendering model to generate a first three-dimensional digital human image; and projecting the first three-dimensional digital human image onto a two-dimensional plane to generate a first expression image.

[0054] After obtaining the first two-dimensional expression image, it is still necessary to determine whether the first expression image is an abnormal expression image.

[0055] Step 103: Input the first facial expression image into the pre-trained evaluation model to obtain the second facial expression parameters.

[0056] The evaluation model is obtained through supervised training using a training set consisting of facial expression sample parameters and expression sample images. This evaluation model can learn the mapping relationship between the input expression sample images and the facial expression sample parameters.

[0057] It should be noted that the second facial expression parameters are different from the first facial expression parameters. The first facial expression parameters are obtained by processing facial driving information, while the second facial expression parameters are obtained by evaluating the first expression image.

[0058] The types of second facial expression parameters are similar to those of the first facial expression parameters, and may include, for example, eye movement parameters, eyelid movement parameters, lip movement parameters, etc.

[0059] Step 104: Process the second facial expression parameters to generate the corresponding second expression image.

[0060] In this step, the rendering model can be the same as the rendering model mentioned in step 102.

[0061] Specifically, step 104 includes: inputting the second facial expression parameters corresponding to each frame of expression into the rendering model to generate a second three-dimensional digital human image; and projecting the second three-dimensional digital human image onto a two-dimensional plane to generate the second expression image.

[0062] In this step, a second three-dimensional digital human image is first rendered based on the second facial expression parameters. Then, the second three-dimensional digital human image is projected onto a two-dimensional plane to generate a second expression image, thereby obtaining an expression image based on the facial expression parameters more objectively.

[0063] Step 105: Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression.

[0064] There are several methods for calculating the similarity between the first expression image and the second expression image, such as calculating cosine similarity and structural similarity measure (SSIM).

[0065] If the similarity is less than the threshold, it indicates that the first expression image and the second expression image are significantly different, such as significant differences in mouth shape or eyelids, and the expression in that frame is marked as an abnormal expression.

[0066] The abnormal expression recognition method based on digital humans provided in this disclosure generates first facial expression parameters corresponding to each frame of expression information based on input facial driving information; processes the first facial expression parameters corresponding to each frame of expression to generate a corresponding first expression image; then inputs the first expression image into a pre-trained evaluation model to obtain second facial expression parameters; processes the second facial expression parameters to generate a corresponding second expression image; calculates the similarity between the first expression image and the second expression image corresponding to each frame of expression; and marks the frame of expression as an abnormal expression if the similarity is less than a threshold. This method can automatically identify abnormal expressions, reduce the burden on testers, and improve processing efficiency.

[0067] Specifically, step 101 includes: inputting the facial driving information into a pre-trained facial driving model to generate first facial expression parameters corresponding to each frame of expression information; wherein, the facial driving information includes at least one of image, video, text, and speech, and the facial driving model is trained by sample driving information and sample expression parameters.

[0068] A pre-trained facial driving model can generate corresponding first facial expression parameters based on facial driving information. For example, if the facial driving information includes the text "haha", then the corresponding first facial expression parameters will be generated so that the final generated facial expression is a "laughing" expression.

[0069] Specifically, after step 105, the method further includes: correcting the expression in the frame to obtain corrected expression information, and generating an expression animation of the digital human based on the corrected expression information.

[0070] The correction can be made by having backend personnel input the corresponding correction command, or by correcting abnormal expressions in the current frame based on the expressions in the previous frame and the expressions in the next frame.

[0071] After the corrections are complete, generate facial expression animations for the digital human, allowing the digital human to perform a series of facial expressions based on the expression information.

[0072] This disclosure also provides a method for recognizing abnormal facial expressions in digital humans. See [link to relevant documentation]. Figure 2 Taking the input facial driving information including text paragraphs as an example, the method includes:

[0073] Step 201: Input the obtained facial driving information into the pre-trained facial driving model to generate the first facial expression parameters corresponding to each frame of expression information.

[0074] Among them, the facial driving information includes timestamp information. The text paragraph can be divided into 1000 frames using the timestamp information, and the first facial expression parameters corresponding to the expressions in the 1000 frames can be generated by the facial driving model.

[0075] Step 202: Input the first facial expression parameters corresponding to each frame of expression into the rendering model to generate the first three-dimensional digital human image.

[0076] Step 203: Project the first three-dimensional digital human image onto a two-dimensional plane to generate the first expression image.

[0077] Step 204: Input the first facial expression image into the pre-trained evaluation model to obtain the second facial expression parameters.

[0078] The evaluation model is obtained through supervised training using a training set consisting of facial expression sample parameters and expression sample images.

[0079] Step 205: Input the second facial expression parameters corresponding to each frame of expression into the rendering model to generate the second three-dimensional digital human image.

[0080] Step 206: Project the second three-dimensional digital human image onto a two-dimensional plane to generate the second expression image.

[0081] Step 207: Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression.

[0082] Ultimately, frame 177 was determined to be an abnormal expression; see [link / reference]. Figure 3 and Figure 4 , Figure 3 and Figure 4 The first and second facial expression images corresponding to the expression in frame 177 are shown respectively. As can be seen from the images, the mouth expressions in the two images are significantly different, which is indeed an abnormal expression.

[0083] Step 208: Correct the expression in the frame to obtain the corrected expression information, and generate the digital human's expression animation based on the corrected expression information.

[0084] Steps 201-208 can automatically identify abnormal facial expressions, reducing the burden on testers and improving processing efficiency.

[0085] The abnormal expression recognition device based on digital human provided in this disclosure is described below. The abnormal expression recognition device based on digital human described below and the abnormal expression recognition method based on digital human described above can be referred to in correspondence.

[0086] This disclosure provides an abnormal facial expression recognition device based on a digital human, see [link to documentation]. Figure 5 ,include:

[0087] The expression parameter generation module 501 is used to generate the first facial expression parameters corresponding to each frame of expression in the expression information based on the input facial driving information.

[0088] The first image processing module 502 is used to process the first facial expression parameters corresponding to each frame of expression to generate the corresponding first expression image.

[0089] Evaluation module 503 is used to input the first facial expression image into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained by supervised training using a training set composed of facial expression sample parameters and facial expression sample images;

[0090] The second image processing module 504 is used to process based on the second facial expression parameters to generate a corresponding second expression image;

[0091] The labeling module 505 is used to calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression, and to label the frame of expression as an abnormal expression if the similarity is less than a threshold.

[0092] Optionally, the expression parameter generation module 501 is specifically used to: input the facial driving information into a pre-trained facial driving model to generate first facial expression parameters corresponding to each frame of expression in the expression information; wherein, the facial driving information includes at least one of image, video, text, and speech;

[0093] The facial driving model is trained using sample driving information and sample expression parameters.

[0094] Optionally, the first image processing module 502 is specifically used for:

[0095] The first facial expression parameters corresponding to each frame of expression are input into the rendering model to generate the first three-dimensional digital human image.

[0096] The first three-dimensional digital human image is projected onto a two-dimensional plane to generate the first facial expression image.

[0097] Optionally, the second image processing module 504 is specifically used for:

[0098] The second facial expression parameters corresponding to each frame of expression are input into the rendering model to generate a second three-dimensional digital human image.

[0099] The second three-dimensional digital human image is projected onto a two-dimensional plane to generate the second facial expression image.

[0100] Optionally, the device further includes: a correction module, configured to correct the expression in the frame after determining that the expression in the frame is an abnormal expression, obtain the corrected expression information, and generate an expression animation of the digital human based on the corrected expression information.

[0101] The abnormal expression recognition device based on digital humans provided in this embodiment generates first facial expression parameters corresponding to each frame of expression information based on input facial driving information; processes the first facial expression parameters corresponding to each frame of expression to generate a corresponding first expression image; then inputs the first expression image into a pre-trained evaluation model to obtain second facial expression parameters; processes the second facial expression parameters to generate a corresponding second expression image; calculates the similarity between the first expression image and the second expression image corresponding to each frame of expression; and marks the frame of expression as an abnormal expression if the similarity is less than a threshold. This allows for automatic identification of abnormal expressions, reducing the burden on testers and improving processing efficiency.

[0102] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a digital human-based abnormal facial expression recognition method, including:

[0103] Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of expression in the expression information;

[0104] The first facial expression parameters corresponding to each frame of expression are processed to generate the corresponding first expression image.

[0105] The first facial expression image is input into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images;

[0106] The second facial expression parameters are processed to generate the corresponding second expression image;

[0107] Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression.

[0108] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0109] On the other hand, this disclosure also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the abnormal facial expression recognition method based on digital humans provided by the above methods, including:

[0110] Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of expression in the expression information;

[0111] The first facial expression parameters corresponding to each frame of expression are processed to generate the corresponding first expression image.

[0112] The first facial expression image is input into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images;

[0113] The second facial expression parameters are processed to generate the corresponding second expression image;

[0114] Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression.

[0115] In another aspect, embodiments of this disclosure also provide a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a digital human-based abnormal facial expression recognition method, including:

[0116] Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of expression in the expression information;

[0117] The first facial expression parameters corresponding to each frame of expression are processed to generate the corresponding first expression image.

[0118] The first facial expression image is input into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images;

[0119] The second facial expression parameters are processed to generate the corresponding second expression image;

[0120] Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression.

[0121] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0122] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit them. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure.

Claims

1. A method for recognizing abnormal facial expressions based on digital humans, characterized in that, include: Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of expression in the expression information; The first facial expression parameters corresponding to each frame of expression are processed to generate the corresponding first expression image. The first facial expression image is input into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images; The second facial expression parameters are processed to generate the corresponding second expression image; Calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression. If the similarity is less than a threshold, mark the frame of expression as an abnormal expression. The first facial expression image is generated by processing the first facial expression parameters corresponding to each frame of expression, including: inputting the first facial expression parameters corresponding to each frame of expression into a rendering model to generate a first three-dimensional digital human image; and projecting the first three-dimensional digital human image onto a two-dimensional plane to generate the first expression image. The process of generating a corresponding second expression image based on the second facial expression parameters includes: inputting the second facial expression parameters corresponding to each frame of expression into a rendering model to generate a second three-dimensional digital human image; and projecting the second three-dimensional digital human image onto a two-dimensional plane to generate the second expression image.

2. The abnormal facial expression recognition method based on digital humans according to claim 1, characterized in that, Based on the input facial driving information, generate the first facial expression parameters corresponding to each frame of the expression information, including: The facial driving information is input into a pre-trained facial driving model to generate first facial expression parameters corresponding to each frame of expression information; wherein, the facial driving information includes at least one of image, video, text, and speech; The facial driving model is trained using sample driving information and sample expression parameters.

3. The abnormal facial expression recognition method based on digital humans according to claim 1, characterized in that, After determining that the expression in the frame is an abnormal expression, the method further includes: The expression in the frame is corrected to obtain the corrected expression information, and the digital human's expression animation is generated based on the corrected expression information.

4. A device for recognizing abnormal facial expressions based on digital humans, characterized in that, include: The expression parameter generation module is used to generate the first facial expression parameters corresponding to each frame of expression information based on the input facial driving information. The first image processing module is used to process the first facial expression parameters corresponding to each frame of expression to generate the corresponding first expression image. An evaluation module is used to input the first facial expression image into a pre-trained evaluation model to obtain the second facial expression parameters; wherein, the evaluation model is obtained through supervised training using a training set composed of facial expression sample parameters and facial expression sample images; The second image processing module is used to process the second facial expression parameters and generate the corresponding second expression image. The labeling module is used to calculate the similarity between the first expression image and the second expression image corresponding to each frame of expression, and to label the frame of expression as an abnormal expression if the similarity is less than a threshold. The first image processing module is specifically used for: inputting the first facial expression parameters corresponding to each frame of expression into the rendering model to generate a first three-dimensional digital human image; projecting the first three-dimensional digital human image onto a two-dimensional plane to generate the first expression image; The second image processing module is specifically used to: input the second facial expression parameters corresponding to each frame of expression into the rendering model to generate a second three-dimensional digital human image; and project the second three-dimensional digital human image onto a two-dimensional plane to generate the second expression image.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the abnormal facial expression recognition method based on digital human as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the abnormal facial expression recognition method based on digital humans as described in any one of claims 1 to 3.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the abnormal facial expression recognition method based on digital humans as described in any one of claims 1 to 3.

Images