A simulated interview method based on AI big model

Abstract

The present invention discloses a simulated interview method based on an AI big model, which relates to the technical field of simulated interviews. The present invention collects the recorded data uploaded by the authorized user after each simulated interview, and analyzes the data by a fuzzification processing module. During the analysis, the recorded video data of each question text contained in the recorded data is fuzzified frame by frame, so that the blurring of each frame of image is reversible. On the one hand, the targetedness of the blurring is ensured while the degree of the blurring is guaranteed, and the blurring strength of different pixel points is dynamic. Compared with the traditional blurring algorithm, the reversibility of the blurring is achieved. On the other hand, the mean is calculated for each pixel point in the image, so that the encryption result of each frame of image is unique, which effectively resists brute force cracking, and only the face area is fuzzified. Compared with encrypting the entire recorded data, computing resources and processing time are saved.

Description

Technical Field

[0001] The present invention relates to the technical field of simulated interviews, and in particular to a simulated interview method based on an AI big model. Background Art

[0002] A mock interview is a training method that uses virtual or simulated interview scenarios to assess an interviewee's comprehensive qualities, including their professional knowledge, communication skills, and adaptability. With the development of artificial intelligence (AI) technology, mock interview methods based on large AI models have gradually become mainstream. By recording a video of a simulated interviewee, AI technology is used to analyze facial data in the video. Combined with information such as the interviewee's language expression and body movements, a comprehensive score is generated based on the interviewee's performance. This score is then used for model training to continuously optimize the accuracy of mock interview assessments.

[0003] In the above process, protecting the facial privacy of the mock interviewee is a very important issue that needs attention. In the existing technology, to ensure that the facial privacy of the mock interviewee is not leaked, the entire video is usually encrypted. From the transmission perspective, encrypting the entire video can form a comprehensive protection barrier during the data transmission process, effectively resisting malicious theft and snooping in the transmission link, and ensuring data security. However, in the storage field, other scene information in the mock interview video besides facial data does not involve privacy. Encrypting the entire video will generate a large amount of encryption redundancy, which not only increases computing resource consumption and processing time, but also wastes storage resources and increases storage costs.

[0004] Currently, in the storage field, common facial encryption methods mostly use fuzzy algorithms. Although they can hide faces, the processing is irreversible. As a result, the original facial information cannot be accurately restored during subsequent interview performance analysis and model training. This affects the AI ​​model's recognition and analysis of key information such as the interviewee's facial expressions and micro-movements, thereby reducing the accuracy of simulated interview assessments and the effectiveness of model training.

[0005] How to achieve reversible encryption of facial data while protecting the privacy of mock interviewees to meet the analysis and training needs of the mock interview process is a technical challenge that needs to be solved in this field.

[0006] In order to solve the above problems, the present invention proposes a solution. Summary of the Invention

[0007] The purpose of the present invention is to provide a simulated interview method based on an AI big model in order to solve the problems raised in the above background technology.

[0008] The present invention provides a simulated interview method based on an AI big model, comprising the following steps:

[0009] Step 1: The user-side acquisition module collects the recorded data of the authorized user's simulated interview and transmits it to the fuzzy processing module. The recorded data contains question numbers of several question texts and recorded video data;

[0010] Step 2: After receiving the transmitted recording data of the simulated interview of the authorized user, the fuzzy processing module processes the recorded video data of each question text contained in the recorded data frame by frame according to the preset fuzzy processing rules to obtain the recorded fuzzy video data of each question text;

[0011] Step 3: After processing the recorded fuzzy video data of each question text contained in the recorded data, the fuzzy processing module generates the recorded fuzzy data of the authorized user's simulated interview and transmits it to the video storage module for storage.

[0012] Furthermore, in step 1, the recorded video data of a question text is a dynamic image stream obtained by recording the authorized user when the authorized user starts to answer the question text verbally.

[0013] Furthermore, the fuzzy processing module includes user information of several authorized users, and the user information includes a registered account, a registered password, a facial image, and the number of simulated interviews.

[0014] Furthermore, in step 2, the fuzzification processing rules for obtaining the fuzzy recorded video data of each question text contained in the recorded data frame by frame are as follows:

[0015] S11: Mark each question text in the recorded data as A1, A2, ..., Aa, a≥1 in ascending order according to the question number;

[0016] S12: extracting recorded video data of the question text A1 from the recorded data, and then marking each frame image contained in the recorded video data as B1, B2, ..., Bb in the order of playback, where b≥1, and b is the total number of frame images contained in the recorded video data;

[0017] b as the authorized user of this simulation related to the breadth of the problem text A1 imprint, re-labeled as C1;

[0018] S13: extracting the number of simulated interviews from the user information of the authorized user stored in the fuzzy processing module, and using the number of simulated interviews as the depth impression amount C2 of the authorized user's current simulation;

[0019] S14: extracting a facial image from the user information of the authorized user stored in the fuzzification processing module, and using the image as a fuzzy reference image of the authorized user for this simulation;

[0020] S15: Determine a magnification E1 of the blurred reference image relative to the question text A1;

[0021] S16: Enlarging the fuzzy reference image according to the magnification factor E1 to obtain a fuzzy enlarged image of the authorized user's current simulation related to the question text A1;

[0022] S17: Mark all pixels constituting the fuzzy enlarged image as F1, F2, ..., Ff in sequence, where f=d1*(x1-1)+y1, where x1 and y1 are the row and column coordinates of each pixel in the fuzzy enlarged image, d1 is the image width of the fuzzy enlarged image, and the value ranges of x1 and y1 are d2 and d1, respectively, and d2 is the image height of the fuzzy enlarged image;

[0023] Mark all the pixels constituting the face area as G1, G2, ..., Gg in sequence;

[0024] S18: reversibly blurring the pixel point G1 in the face area of ​​the image B1 according to a preset blurring rule;

[0025] S19: reversibly blurring the pixel points G2, G3, ..., Gg in the face region of the image B1 in sequence according to S18, and calibrating the image B1 after reversibly blurring the pixel point Gg as a blurred image of the image B1 before reversible blurring;

[0026] S110: Obtain blurred images of images B2, B3, ..., Bb before reversible blurring in sequence according to S17 to S19; combine the blurred images of images B1, B2, ..., Bb frame by frame in ascending order of their marking subscripts to obtain recorded blurred video data of question text A1;

[0027] S111: Obtain recorded fuzzy video data of question texts A2, A3, ..., Aa in sequence according to S11 to S110.

[0028] Furthermore, S15, the magnification E1 is determined as follows:

[0029] S151: performing face recognition on the image B1, selecting a face region in the image B1, and counting the total number D1 of pixels constituting the face region in the image B1;

[0030] S152: Calculate the magnification E1 of the blurred reference image relative to the question text A1 using the formula E1=ceil(D1*C1*P1 / P2), where P1 is a preset blurred pixel scalar, P2 is the total number of pixels constituting the blurred reference image, and ceil() is a round-up function.

[0031] Furthermore, in S17, g=d3*(x2-1)+y2, where x2 and y2 are the row and column coordinates of the pixel coordinates of each pixel point in the face area, d3 is the image width of the face area, and the value ranges of x2 and y2 are d4 and d3, respectively, and d4 is the image height of the face area.

[0032] Furthermore, in S18, the fuzzy rule for reversibly blurring the pixel G1 is as follows:

[0033] S181: Based on P1, all pixels with a marking subscript less than or equal to P1-1 are extracted from the pixels F1, F2, ..., Ff, that is, the pixels F1, F2, ..., FP1-1 are extracted, and the pixel values ​​H1, H2, ..., HP1-1 of the pixels F1, F2, ..., FP1-1 under the R color component are sequentially obtained from the fuzzy magnified image based on the pixels F1, F2, ..., FP1-1;

[0034] S182: Obtaining a pixel value I1 of the pixel point G1 in the R color component in the face area;

[0035] S183: Calculate the average value J1 of the sum of the pixel values ​​H1, H2, ..., HP1-1, and I1 using an addition and averaging formula;

[0036] S184: Compare J1 and floor(255 / 2)-C2-1. Floor() is a floor rounding function. If J1 ≤ floor(255 / 2)-C2, then the value of J1+C2 is used as the fuzzy pixel value of the pixel point G1 under the R color component. Otherwise, the value of J1-C2 is used as the fuzzy pixel value of the pixel point G1 under the R color component. The 1 in "-1" is the 1 in the mark subscript of the question text A1.

[0037] S185: Calculate and obtain the fuzzy pixel value of the pixel point G1 under the G color component and the B color component in sequence according to S181 to S184;

[0038] S186: Replace the pixel values ​​of the pixel point G1 in the face area of ​​the image B1 under the R, G, and B color components according to the fuzzy pixel values ​​of the pixel point G1 under R, G, and B.

[0039] Furthermore, in step three, the fuzzy processing module transmits the recorded fuzzy data of the simulated interview of the authorized user to the video storage module for storage and simultaneously increases the number of simulated interviews in the user information of the authorized user stored therein by 1.

[0040] Compared with the existing technology, it has the following beneficial effects:

[0041] The present invention collects the recorded data uploaded by the authorized user after each simulated interview, and analyzes it by a fuzzification processing module. During the analysis process, the recorded video data of each question text contained in the recorded data is fuzzified frame by frame. During the fuzzification process, the facial capture image of the authorized user and the number of simulated interviews are combined as the basis for fuzzification of the pixels constituting the face area in each frame image, so that the fuzzification of each frame image is reversible. In the process of fuzzification of each frame image, the total number of frame images in the recorded video data corresponding to the question text is used as the breadth imprint of the corresponding question text to determine the magnification of the facial capture image for different question texts and amplify the facial capture image. Based on the amplified pixel points The mean value of the pixels that make up the facial image in the image is calculated. On the one hand, it ensures the targetedness of the blurring while ensuring the degree of blurring, making the blurring strength of different pixels dynamic, and realizing the reversibility of blurring compared to traditional blurring algorithms. On the other hand, the pixels for mean calculation of each pixel in each frame of the image are different, which not only makes the encryption result of each frame of the image unique, but also, as the number of interviews increases and the length of the recorded video is different when answering the question text, the blurring strength of each frame of the image is dynamic, further enhancing the dynamic defense capability of encryption, effectively resisting brute force cracking and data reverse analysis, and only blurring the face area, compared with encrypting the entire recorded data, saving computing resources and processing time. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0044] See also Figure 1, this application provides a simulated interview method based on AI big model, the above method is performed by a simulated interview system based on AI big model, the system includes a user side acquisition module, a fuzzy processing module and a video storage module;

[0045] The user-side collection module is used to collect the recorded data of the simulated interview uploaded by the authorized user after each simulated interview, where the recorded data includes a number of question texts and question numbers and recorded video data;

[0046] In this application, an authorized user refers to a logged-in user who has been authenticated. The recorded video data of a question text is a dynamic image stream obtained by recording the authorized user when the authorized user begins to answer the question text. The video length of the recorded video data of a question text is from the time when the authorized user begins to answer the question text to the time when the authorized user completes the answer to the question text.

[0047] In this application, for any question text, the authorized user is provided with a start answering button and an end answering button. When the authorized user clicks the start answering button, it means that the authorized user starts answering the current question according to the question text, and clicks the end answering button to indicate that the authorized user has finished answering the question text in person;

[0048] In this application, the question texts are numbered starting from 1 and continuing in sequence to identify the order in which the authorized users answer the corresponding question texts;

[0049] In this application, the question numbers of the question texts in the recorded data of any simulated interview are all numbered starting from 1 and continuing forward;

[0050] In this application, one of the recorded video data contains the facial expressions, body movements and spatial postures of the authorized user;

[0051] After the mock interview is over, the authorized user uploads the recorded data of the mock interview. The user-side acquisition module acquires the recorded data of the authorized user's mock interview and transmits it to the fuzzy processing module.

[0052] A fuzzy processing module is used to fuzzy the recorded data of each simulated interview of the authorized user. The fuzzy processing module contains user information of several authorized users, and the user information includes a registered account, a registered password, a facial image, and the number of simulated interviews. In this application, the facial image is the image data obtained by capturing the authorized user's face when the authorized user registers the account;

[0053] The fuzzy processing module receives the transmitted recording data of the simulated interview of the authorized user and processes the recorded video data of each question text contained in the recorded data frame by frame according to the preset fuzzy processing rules to obtain the recorded fuzzy video data of each question text. The fuzzy processing rules are as follows:

[0054] S11: Mark each question text in the recorded data as A1, A2, ..., Aa, a≥1 in ascending order according to the question number;

[0055] S12: extracting recorded video data of the question text A1 from the recorded data, and then marking each frame image contained in the recorded video data as B1, B2, ..., Bb in the order of playback, where b≥1, and b is the total number of frame images contained in the recorded video data;

[0056] b as the authorized user of this simulation related to the breadth of the problem text A1 imprint, re-labeled as C1;

[0057] S13: extracting the number of simulated interviews from the user information of the authorized user stored in the fuzzy processing module, and using the number of simulated interviews as the depth impression amount C2 of the authorized user's current simulation;

[0058] S14: extracting a facial image from the user information of the authorized user stored in the fuzzification processing module, and using the image as a fuzzy reference image of the authorized user for this simulation;

[0059] S15: Determine the magnification E1 of the blurred reference image relative to the question text A1 according to a preset determination rule. The determination rule is as follows:

[0060] S151: performing face recognition on the image B1, selecting a face region in the image B1, and counting the total number D1 of pixels constituting the face region in the image B1;

[0061] The facial region is selected by using the Viola-Jones algorithm with the help of OpenCV software using Haar features and cascade classifiers to quickly screen the facial region;

[0062] S152: Calculate the magnification E1 of the blurred reference image relative to the question text A1 using the formula E1=ceil(D1*C1*P1 / P2), where P1 is a preset blurred pixel scalar, whose value is set by the administrator based on the available computing resources of the system. The value of P1 is an even number. In this application, the value of P1 is 4. P2 is the total number of pixels constituting the blurred reference image, and ceil() is a round-up function.

[0063] S16: Enlarging the fuzzy reference image according to the magnification factor E1 to obtain a fuzzy enlarged image of the authorized user's current simulation related to the question text A1;

[0064] S17: Mark all pixels constituting the fuzzy enlarged image as F1, F2, ..., Ff in sequence, where f=d1*(x1-1)+y1, where x1 and y1 are the row and column coordinates of each pixel in the fuzzy enlarged image, d1 is the image width of the fuzzy enlarged image, and the value ranges of x1 and y1 are d2 and d1, respectively, and d2 is the image height of the fuzzy enlarged image;

[0065] Mark all the pixels constituting the face area as G1, G2, ..., Gg, where g = d3*(x2-1)+y2, where x2 and y2 are the row and column coordinates of each pixel in the face area, d3 is the image width of the face area, and the value ranges of x2 and y2 are d4 and d3, respectively, where d4 is the image height of the face area.

[0066] S18: reversibly blur pixel G1 in the face region of image B1 according to a preset fuzzy rule. The fuzzy rule is as follows:

[0067] S181: Based on P1, extract all pixels whose marking subscripts are less than or equal to P1-1 from the pixels F1, F2, ..., Ff, that is, extract the pixels F1, F2, ..., FP1-1;

[0068] sequentially acquiring pixel values ​​H1, H2, ..., HP1-1 of the pixel points F1, F2, ..., FP1-1 under the R color component from the blurred magnified image according to the pixel points F1, F2, ..., FP1-1;

[0069] S182: Obtaining a pixel value I1 of the pixel point G1 in the R color component in the face area;

[0070] S183: Calculate the average value J1 of the sum of the pixel values ​​H1, H2, ..., HP1-1, and I1 using an addition and averaging formula;

[0071] S184: Compare J1 and floor(255 / 2)-C2-1. Floor() is a floor rounding function. If J1 ≤ floor(255 / 2)-C2, then the value of J1+C2 is used as the fuzzy pixel value of the pixel point G1 under the R color component. Otherwise, the value of J1-C2 is used as the fuzzy pixel value of the pixel point G1 under the R color component. The 1 in "-1" is the 1 in the mark subscript of the question text A1.

[0072] S185: Calculate and obtain the fuzzy pixel value of the pixel point G1 under the G color component and the B color component in sequence according to S181 to S184;

[0073] S186: replacing the pixel values ​​of the pixel point G1 in the face region of the image B1 under R, G, and B according to the fuzzy pixel values ​​of the pixel point G1 under R, G, and B;

[0074] S19: reversibly blurring pixel points G2, G3, ..., Gg in the face region of image B1 in sequence according to S18, and calibrating image B1 after reversibly blurring pixel point Gg as a blurred image of image B1 before reversible blurring. It should be noted that the face region has been framed in the blurred image;

[0075] S110: Obtain blurred images of images B2, B3, ..., Bb before reversible blurring in sequence according to S17 to S19; combine the blurred images of images B1, B2, ..., Bb frame by frame in ascending order of their marking subscripts to obtain recorded blurred video data of question text A1;

[0076] S111: Obtain recorded fuzzy video data of question texts A2, A3, ..., Aa in sequence according to S11 to S110, wherein when calculating the fuzzy pixel value of a pixel point in the face area of ​​any frame image in the recorded mode video data of question text A2 under R, G, and B color components, it is floor(255 / 2)-C2-2, where the 2 in "-2" is the 2 of the mark subscript of question text A2, and calculate the fuzzy pixel value of a pixel point in the face area of ​​any frame image in the recorded mode video data of other question texts under R, G, and B color components, and so on;

[0077] It should be noted here that the R, G, and B color components are the three color components in the RGB color mode;

[0078] The fuzzy processing module generates the recorded fuzzy data of the simulated interview of the authorized user according to the recorded fuzzy video data of each question text contained in the recorded data of the simulated interview of the authorized user, and transmits the recorded fuzzy data to the video storage module;

[0079] At the same time, the fuzzy processing module increases the number of simulated interviews in the user information of the authorized user stored therein by 1;

[0080] The video storage module is used to store the recorded fuzzy data of each simulated interview of several authorized users. After receiving the transmitted recorded fuzzy data of the simulated interview of the authorized user, the video storage module stores it to facilitate scoring the simulated interview of the authorized user and analyzing the recorded fuzzy data of the simulated interview of the authorized user for model training;

[0081] After receiving the video analysis instruction input by the administrator, the video storage module extracts the recorded fuzzy data of all simulated interviews of all authorized users within the time range from the video storage module according to the time range carried in the video analysis instruction;

[0082] For the extracted recorded fuzzy data of any mock interview of any authorized user, the number of mock interviews when generating the recorded fuzzy data of the mock interview of the authorized user is calculated based on the order in which the recorded fuzzy data of each mock interview of the authorized user stored in the video storage module and the number of mock interviews in the user information of the authorized user stored in the current fuzzification processing module;

[0083] Restore the recorded fuzzy video data of each question text in the recorded fuzzy data by inversely executing the fuzzification processing rules according to the question numbers in ascending order, combined with the number of simulated interviews and the facial image of the authorized user, to obtain the recorded data of the simulated interview of the authorized user; temporarily store the recorded data for a period determined by the analysis requirements of the management personnel, where the analysis requirements include interview scoring and model training;

[0084] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0085] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A simulated interview method based on AI big model, characterized by: The following steps are involved: Step 1: The user-side acquisition module collects the recorded data of the authorized user's simulated interview and transmits it to the fuzzy processing module. The recorded data contains question numbers of several question texts and recorded video data; Step 2: After receiving the transmitted recording data of the simulated interview of the authorized user, the fuzzification processing module processes the recorded video data of each question text contained in the recorded data frame by frame according to the preset fuzzification processing rules to obtain the recorded fuzzy video data of each question text. The fuzzification processing rules are as follows: S11: Mark each question text in the recorded data as A1, A2, ..., Aa, a≥1 in ascending order according to the question number; S12: extracting recorded video data of the question text A1 from the recorded data, and then marking each frame image contained in the recorded video data as B1, B2, ..., Bb in the order of playback, where b≥1, and b is the total number of frame images contained in the recorded video data; b as the authorized user of this simulation related to the breadth of the problem text A1 imprint, re-labeled as C1; S13: extracting the number of simulated interviews from the user information of the authorized user stored in the fuzzy processing module, and using the number of simulated interviews as the depth impression amount C2 of the authorized user's current simulation; S14: extracting a facial image from the user information of the authorized user stored in the fuzzification processing module, and using the image as a fuzzy reference image of the authorized user for this simulation; S15: Determine a magnification E1 of the blurred reference image relative to the question text A1; S16: Enlarging the fuzzy reference image according to the magnification factor E1 to obtain a fuzzy enlarged image of the authorized user's current simulation related to the question text A1; S17: Mark all pixels constituting the fuzzy enlarged image as F1, F2, ..., Ff in sequence, where f=d1*(x1-1)+y1, where x1 and y1 are the row and column coordinates of each pixel in the fuzzy enlarged image, d1 is the image width of the fuzzy enlarged image, and the value ranges of x1 and y1 are d2 and d1, respectively, and d2 is the image height of the fuzzy enlarged image; Mark all the pixels constituting the face area as G1, G2, ..., Gg in sequence; S18: reversibly blurring the pixel point G1 in the face area of ​​the image B1 according to a preset blurring rule; S19: reversibly blurring the pixel points G2, G3, ..., Gg in the face region of the image B1 in sequence according to S18, and calibrating the image B1 after reversibly blurring the pixel point Gg as a blurred image of the image B1 before reversible blurring; S110: Obtain blurred images of images B2, B3, ..., Bb before reversible blurring in sequence according to S17 to S19; combine the blurred images of images B1, B2, ..., Bb frame by frame in ascending order of their marking subscripts to obtain recorded blurred video data of question text A1; S111: Obtain recorded fuzzy video data of question texts A2, A3, ..., Aa in sequence according to S11 to S110; Step 3: After the fuzzy video data of each question text contained in the recorded data is processed, the fuzzy processing module generates the fuzzy data of the simulated interview of the authorized user and transmits it to the video storage module for storage; The fuzzy processing module contains user information of several authorized users, and the user information includes a registered account, a registered password, a facial image and the number of simulated interviews.

2. A simulated interview method based on AI big model according to claim 1, characterized in that: In step 1, the recorded video data of a question text is a dynamic image stream obtained by recording the authorized user when the authorized user starts to answer the question text verbally.

3. A simulated interview method based on AI big model according to claim 1, characterized in that: S15, the rules for determining the magnification E1 are as follows: S151: performing face recognition on the image B1, selecting a face region in the image B1, and counting the total number D1 of pixels constituting the face region in the image B1; S152: Calculate the magnification E1 of the blurred reference image relative to the question text A1 using the formula E1=ceil(D1*C1*P1 / P2), where P1 is a preset blurred pixel scalar, P2 is the total number of pixels constituting the blurred reference image, and ceil() is a round-up function.

4. The simulated interview method based on AI big model according to claim 1 is characterized in that: In S17, g=d3*(x2-1)+y2, where x2 and y2 are the row and column coordinates of the pixel coordinates of each pixel point in the face area, d3 is the image width of the face area, and the value ranges of x2 and y2 are d4 and d3, respectively, where d4 is the image height of the face area.

5. The simulated interview method based on AI big model according to claim 1 is characterized in that: S18, the fuzzy rule for reversibly blurring pixel G1 is as follows: S181: Based on P1, all pixels with a marking subscript less than or equal to P1-1 are extracted from the pixels F1, F2, ..., Ff, that is, the pixels F1, F2, ..., FP1-1 are extracted, and the pixel values ​​H1, H2, ..., HP1-1 of the pixels F1, F2, ..., FP1-1 under the R color component are sequentially obtained from the fuzzy magnified image based on the pixels F1, F2, ..., FP1-1; S182: Obtaining a pixel value I1 of the pixel point G1 in the R color component in the face area; S183: Calculate the average value J1 of the sum of the pixel values ​​H1, H2, ..., HP1-1, and I1 using an addition and averaging formula; S184: Compare J1 and floor(255 / 2)-C2-1. Floor() is a floor rounding function. If J1 ≤ floor(255 / 2)-C2, then the value of J1+C2 is used as the fuzzy pixel value of the pixel point G1 under the R color component. Otherwise, the value of J1-C2 is used as the fuzzy pixel value of the pixel point G1 under the R color component. The 1 in "-1" is the 1 in the mark subscript of the question text A1. S185: Calculate and obtain the fuzzy pixel value of the pixel point G1 under the G color component and the B color component in sequence according to S181 to S184; S186: Replace the pixel values ​​of the pixel point G1 in the face area of ​​the image B1 under the R, G, and B color components according to the fuzzy pixel values ​​of the pixel point G1 under R, G, and B.

6. The simulated interview method based on AI big model according to claim 1 is characterized in that: In step three, when the fuzzy processing module transmits the recorded fuzzy data of the authorized user's simulated interview to the video storage module for storage, the number of simulated interviews in the user information of the authorized user stored therein is simultaneously increased by 1.

Images