Risk assessment method and device, computer equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, relates to a risk assessment method and device, computer equipment and a storage medium, and is applied to the field of digital medical treatment. Performing face overall feature extraction on the face video to obtain a face overall feature vector; performing face independent feature extraction on the face video to obtain a face independent feature coding vector; performing face combination feature extraction on the face video to obtain a face combination feature coding vector; performing fusion processing on the face overall feature vector, the face independent feature coding vector and the face combined feature coding vector to obtain a face feature vector; and performing risk assessment on the facial feature vector based on a risk assessment model to obtain a risk assessment result of the user. In addition, the risk assessment result can be stored in the block chain. Based on the use of the risk assessment model, the processing efficiency and the processing reliability of the risk assessment of the mental health are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of digital medicine, and in particular to risk assessment methods, devices, computer equipment and storage media. Background Art

[0002] In today's rapidly developing modern society, the importance of mental health is increasingly gaining widespread attention and recognition. This shift stems from people's gradual recognition that mental health is not only a key component of personal well-being, but also an important cornerstone of social stability and harmony. Historically, due to cultural concepts, cognitive limitations and social misunderstandings of mental problems, people tend to ignore mental health issues, or even regard them as a shame or personal weakness, and are reluctant to discuss them openly or seek help.

[0003] However, with the acceleration of globalization, the rapid acceleration of the pace of life, and the continuous accumulation of pressure from work, study, family, etc., mental health issues have become increasingly prominent and have become an important factor affecting the quality of life and happiness of individuals. The incidence of mental illness has increased significantly, which not only imposes a huge physical and mental burden on individuals, but also has an unignorable impact on social and economic development. Therefore, it has become an urgent task to raise public awareness of mental health issues and strengthen the construction of a mental health service system.

[0004] In the practice of mental health services, timely detection and effective intervention of mental health problems are crucial. Early identification of mental health problems can not only alleviate the individual's pain and improve the quality of life, but also effectively prevent the further deterioration of mental problems and reduce social costs. In order to achieve this goal, scale surveys are currently widely used to assess the risk of users' mental health. These scales usually contain a series of questions designed to assess an individual's mental state, emotional response, behavioral patterns, etc., thereby providing a quantitative basis for their mental health status.

[0005] However, although scale surveys play an important role in mental health risk assessment, their reliability is subject to certain limitations. Some users may conceal their true situation when filling out the scale for various reasons (such as fear of being discriminated against, fear of being misunderstood, and protection of personal privacy). This concealment behavior will not only greatly reduce the accuracy of the scale survey results, but may also make subsequent mental health intervention measures lack of pertinence and difficult to achieve the expected results. Therefore, how to improve the reliability of mental health risk assessment and ensure that the scale survey can truly reflect the user's mental health status is an important issue that needs to be urgently addressed in the current mental health service field. Summary of the invention

[0006] The purpose of the embodiments of the present application is to propose a risk assessment method, apparatus, computer equipment and storage medium to solve the technical problem of low reliability of existing mental health risk assessment methods based on scale surveys.

[0007] In order to solve the above technical problems, the present application embodiment provides a risk assessment method, which adopts the following technical solution:

[0008] Get the user's facial video;

[0009] Performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector;

[0010] Performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector;

[0011] Performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector;

[0012] The facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector are fused to obtain a corresponding facial feature vector;

[0013] The facial feature vector is subjected to risk assessment processing based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

[0014] Furthermore, the step of performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector specifically includes:

[0015] Preprocessing the facial video to obtain a corresponding designated facial video;

[0016] Call the preset facial overall feature extraction model;

[0017] Performing vector coding processing on the designated facial video based on the facial overall feature extraction model to obtain a corresponding vector coding result;

[0018] The vector encoding result is used as the overall facial feature vector.

[0019] Furthermore, the step of performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector specifically includes:

[0020] Splitting the facial video into a corresponding plurality of video frames;

[0021] Call the preset facial independent feature extraction model;

[0022] Performing facial independent feature extraction processing on all the video frames based on the facial independent feature extraction model to obtain corresponding first feature vectors;

[0023] Based on the time sequence of the video frames, all the first feature vectors are concatenated to obtain a corresponding first feature matrix;

[0024] The first feature matrix is ​​encoded based on a preset first encoding model to obtain the facial independent feature encoding vector.

[0025] Furthermore, the step of performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector specifically includes:

[0026] Splitting the facial video into a corresponding plurality of designated videos based on a preset sliding time window;

[0027] Call the preset facial combination feature extraction model;

[0028] Performing facial combination feature extraction processing on all the specified videos based on the facial combination feature extraction model to obtain corresponding second feature vectors;

[0029] Based on the time sequence of the specified video, all the second feature vectors are concatenated to obtain a corresponding second feature matrix;

[0030] The second feature matrix is ​​encoded based on a preset second encoding model to obtain the facial combination feature encoding vector.

[0031] Furthermore, the step of fusing the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector to obtain a corresponding facial feature vector specifically includes:

[0032] Get the preset vector splicing strategy;

[0033] Based on the vector splicing strategy, vector splicing processing is performed on the facial overall feature vector, the facial independent feature coding vector, and the facial combined feature coding vector to obtain a corresponding splicing vector;

[0034] The concatenated vector is used as the facial feature vector.

[0035] Furthermore, before the step of performing risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user, the method further includes:

[0036] Get access to a sample set of pre-built risk assessments;

[0037] Dividing the risk assessment sample set into a training set and a validation set based on a preset division ratio;

[0038] Call the preset deep learning model;

[0039] Training the deep learning model based on the training set to obtain a trained first deep learning model;

[0040] Performing a performance evaluation on the first deep learning model based on the validation set to obtain a corresponding performance evaluation result;

[0041] Optimizing the first deep learning model based on the performance evaluation result to obtain a second deep learning model that meets preset performance expectations;

[0042] The second deep learning model is used as the risk assessment model.

[0043] Furthermore, after the step of performing risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user, the method further includes:

[0044] Determine whether the risk assessment results meet the preset risk conditions;

[0045] If so, generating corresponding risk warning information based on the risk assessment result;

[0046] Obtaining an information push method corresponding to the user;

[0047] Based on the information push method, the risk warning information is sent to the user.

[0048] In order to solve the above technical problems, the embodiment of the present application further provides a risk assessment device, which adopts the following technical solution:

[0049] A first acquisition module is used to acquire a user's facial video;

[0050] A first extraction module is used to perform facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector;

[0051] A second extraction module is used to perform facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector;

[0052] A third extraction module is used to perform facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector;

[0053] A fusion module, used for fusing the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector to obtain a corresponding facial feature vector;

[0054] An assessment module is used to perform risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

[0055] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0056] Get the user's facial video;

[0057] Performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector;

[0058] Performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector;

[0059] Performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector;

[0060] The facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector are fused to obtain a corresponding facial feature vector;

[0061] The facial feature vector is subjected to risk assessment processing based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

[0062] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0063] Get the user's facial video;

[0064] Performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector;

[0065] Performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector;

[0066] Performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector;

[0067] The facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector are fused to obtain a corresponding facial feature vector;

[0068] The facial feature vector is subjected to risk assessment processing based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

[0069] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0070] The present application first obtains the user's facial video; then performs facial overall feature extraction processing on the facial video to obtain the corresponding facial overall feature vector; and performs facial independent feature extraction processing on the facial video to obtain the corresponding facial independent feature coding vector; and performs facial combined feature extraction processing on the facial video to obtain the corresponding facial combined feature coding vector; subsequently, the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector are fused to obtain the corresponding facial feature vector; finally, the facial feature vector is subjected to risk assessment processing based on a preset risk assessment model to obtain a risk assessment result corresponding to the user. The present application obtains the user's facial video, then performs facial overall feature extraction, facial independent feature extraction and facial combined feature extraction on the facial video, and fuses the obtained facial overall feature vector, facial independent feature coding vector and facial combined feature coding vector to obtain a facial feature vector, and then performs risk assessment processing on the facial feature vector based on the use of the risk assessment model, so that the risk assessment of the user's mental health can be completed intelligently and accurately and a risk assessment result corresponding to the user is generated, which effectively improves the processing efficiency and accuracy of the risk assessment of mental health and ensures the reliability of the generated risk assessment result. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0072] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0073] Figure 2 A flow chart of an embodiment of a risk assessment method according to the present application;

[0074] Figure 3 is a schematic structural diagram of an embodiment of a risk assessment device according to the present application;

[0075] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0076] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0077] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0078] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0079] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0080] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0081] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0082] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0083] It should be noted that the risk assessment method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the risk assessment device is generally set in the server / terminal device.

[0084] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0085] Continue to refer Figure 2 , showing a flow chart of an embodiment of the risk assessment method according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The risk assessment method provided in the embodiment of the present application can be applied to any scenario that requires risk assessment, and the risk assessment method can be applied to products in these scenarios, for example, mental health assessment in the field of digital medicine. The risk assessment method comprises the following steps:

[0086] Step S201, obtaining a user's facial video.

[0087] In this embodiment, the electronic device (eg, Figure 1The server / terminal device shown in the figure) can obtain the user's facial video through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The executor of the present application may specifically be a risk assessment system, or simply referred to as a system. The present application can be applied to the processing scenarios of risk assessment of mental health in the field of digital medicine. The user's facial video can be obtained by shooting the user's face with a camera during a conversation between a doctor (or system) and a user about their mental health status.

[0088] Step S202, performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector.

[0089] In this embodiment, the above-mentioned specific implementation process of performing facial overall feature extraction processing on the facial video to obtain the corresponding facial overall feature vector will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated in detail here.

[0090] Step S203, performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector.

[0091] In this embodiment, the above-mentioned specific implementation process of performing facial independent feature extraction processing on the facial video to obtain the corresponding facial independent feature encoding vector will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated in detail here.

[0092] Step S204, performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector.

[0093] In this embodiment, the above-mentioned facial combination feature extraction processing is performed on the facial video to obtain the specific implementation process of the corresponding facial combination feature encoding vector. This application will provide further details on this in subsequent specific embodiments and will not be elaborated on here.

[0094] Step S205, fusing the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector to obtain a corresponding facial feature vector.

[0095] In this embodiment, the above-mentioned specific implementation process of fusing the overall facial feature vector, the independent facial feature coding vector and the combined facial feature coding vector to obtain the corresponding facial feature vector will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0096] Step S206, performing risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

[0097] In this embodiment, the facial feature vector can be input into the risk assessment model, and the risk assessment model can be used to perform risk assessment on the facial feature vector, and output a risk assessment result corresponding to the user's mental health (a risk label corresponding to mental health). The risk assessment result may include no risk, low risk, medium risk, and high risk. The model construction process of the risk assessment model will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.

[0098] The present application first obtains the user's facial video; then performs facial overall feature extraction processing on the facial video to obtain the corresponding facial overall feature vector; and performs facial independent feature extraction processing on the facial video to obtain the corresponding facial independent feature coding vector; and performs facial combined feature extraction processing on the facial video to obtain the corresponding facial combined feature coding vector; subsequently, the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector are fused to obtain the corresponding facial feature vector; finally, the facial feature vector is subjected to risk assessment processing based on a preset risk assessment model to obtain a risk assessment result corresponding to the user. The present application obtains the user's facial video, then performs facial overall feature extraction, facial independent feature extraction and facial combined feature extraction on the facial video, and fuses the obtained facial overall feature vector, facial independent feature coding vector and facial combined feature coding vector to obtain a facial feature vector, and then performs risk assessment processing on the facial feature vector based on the use of the risk assessment model, so that the risk assessment of the user's mental health can be completed intelligently and accurately and a risk assessment result corresponding to the user is generated, which effectively improves the processing efficiency and accuracy of the risk assessment of mental health and ensures the reliability of the generated risk assessment result.

[0099] In some optional implementations, step S202 includes the following steps:

[0100] The facial video is preprocessed to obtain a corresponding designated facial video.

[0101] In this embodiment, the acquired facial video may be preprocessed according to the model input requirements of the facial overall feature extraction model, such as adjusting the resolution, frame rate, etc., to obtain the corresponding designated facial video.

[0102] Call the preset facial overall feature extraction model.

[0103] In this embodiment, the network structure of the facial overall feature extraction model is a video vision coding network ViViT (Video Vision Transformer), which has been pre-trained on large-scale video data and has strong video understanding capabilities (i.e., encoding capabilities). In order to enhance the ViViT network's ability to understand facial videos, the pre-trained ViViT network is trained in an unsupervised manner by collecting existing, large amounts of facial videos, thereby obtaining the final facial front feature extraction model.

[0104] The designated facial video is subjected to vector coding processing based on the overall facial feature extraction model to obtain a corresponding vector coding result.

[0105] In this embodiment, the input of the facial overall feature extraction model is the user's facial video. After the facial overall feature extraction model performs vector encoding on the facial video, it will output a corresponding overall feature encoding vector.

[0106] The vector encoding result is used as the overall facial feature vector.

[0107] The present application obtains the corresponding designated facial video by preprocessing the facial video; then calls a preset overall facial feature extraction model; then performs vector encoding processing on the designated facial video based on the overall facial feature extraction model to obtain a corresponding vector encoding result; and subsequently uses the vector encoding result as the overall facial feature vector. The present application obtains the designated facial video by preprocessing the facial video, and then performs vector encoding processing on the designated facial video based on the call of the overall facial feature extraction model, so that the overall facial feature extraction processing of the facial video can be completed quickly and accurately, thereby improving the processing efficiency of the overall facial feature extraction and ensuring the data accuracy of the overall facial feature vector obtained.

[0108] In some optional implementations of this embodiment, step S203 includes the following steps:

[0109] The facial video is split into a corresponding plurality of video frames.

[0110] In this embodiment, the facial video data may be divided into separate frames for subsequent frame-by-frame analysis.

[0111] Call the preset facial independent feature extraction model.

[0112] In this embodiment, the network structure of the above-mentioned facial independent feature extraction model is ViT (Vision Transformer), which has been pre-trained on large-scale image data. Model training is performed by constructing a training data set, the input data in the training data set is a facial image, and the output data in the training data set is the independent features that appear in the facial image mark. By applying the above-mentioned training data set, the pre-trained ViT network is end-to-end supervised model training to obtain the final facial independent feature extraction model.

[0113] Based on the facial independent feature extraction model, facial independent feature extraction processing is performed on all the video frames to obtain corresponding first feature vectors.

[0114] In this embodiment, for the constructed facial independent feature extraction model, the input of the facial independent feature extraction model is each frame of the user's facial video, the facial independent feature extraction model extracts facial independent features from the user's facial picture, and the output of the facial independent feature extraction model is the extraction result of facial independent features, that is, an M-dimensional 0-1 vector. The M dimension is a total of M independent features related to mental health risk assessment, such as raised inner eyebrows, raised chin, upward nasolabial grooves, etc. Each independent feature corresponds to a fixed position in the M-dimensional vector. A value of 0 indicates that the independent feature is recognized, and a value of 1 indicates that the independent feature is not recognized.

[0115] All the first feature vectors are concatenated based on the time sequence of the video frames to obtain a corresponding first feature matrix.

[0116] In this embodiment, after extracting facial independent features from each video frame of the facial video through the above-mentioned facial independent feature extraction model to obtain the corresponding first feature vector, the above-mentioned first feature vectors can be spliced ​​according to the time sequence of the video frames to obtain a first feature matrix, wherein the rows of the first feature matrix correspond to each independent feature, and the columns are each frame of the facial video arranged in time sequence.

[0117] The first feature matrix is ​​encoded based on a preset first encoding model to obtain the facial independent feature encoding vector.

[0118] In this embodiment, the first encoding model may specifically adopt a convolutional neural network (CNN). By using the convolutional neural network to encode the extracted first feature matrix, the required facial independent feature encoding vector may be obtained.

[0119] The present application splits the facial video into corresponding multiple video frames; then calls a preset facial independent feature extraction model; and extracts facial independent features from all the video frames based on the facial independent feature extraction model to obtain a corresponding first feature vector; then splices all the first feature vectors based on the time sequence of the video frames to obtain a corresponding first feature matrix; and then encodes the first feature matrix based on a preset first encoding model to obtain the facial independent feature encoding vector. The present application splits the facial video into corresponding multiple video frames; then extracts facial independent features from all the video frames based on the call of a facial independent feature extraction model to obtain a first feature vector; then splices all the first feature vectors based on the time sequence of the video frames to obtain a first feature matrix; and then encodes the first feature matrix based on the use of the first encoding model, so that the facial independent feature extraction of the facial video can be completed quickly and accurately, the processing efficiency of the facial independent feature extraction is improved, and the data accuracy of the obtained facial independent feature encoding vector is ensured.

[0120] In some optional implementations, step S204 includes the following steps:

[0121] The facial video is split into a corresponding plurality of designated videos based on a preset sliding time window.

[0122] In this embodiment, facial independent features are features extracted from each frame in a facial video. Different from facial independent features, facial composite features set a sliding time window of X seconds to extract facial composite features related to mental health risk assessment from X seconds of a facial video. Facial composite features, such as smile suppression, are described by a continuous change in facial movements, such as upper lip lifting when the cheeks are lifted but not ended normally, and lowering of the corners of the mouth when the corners of the mouth are raised but not ended normally. Among them, a sliding time window of X seconds can be applied to split the facial video to obtain multiple facial videos of X seconds. There is no limitation on the value of X seconds, and it can be set according to actual business needs. For example, X seconds can be set to 2 seconds.

[0123] Call the preset facial combination feature extraction model.

[0124] In this embodiment, the network structure of the facial combination feature extraction model is a video vision encoding network ViViT (Video Vision Transformer), which has been pre-trained on large-scale video data. Training data is constructed for model training, and the input data in the training data is X seconds of user facial video, and the output data in the training data is the combined features of the appearance of the markers of the X seconds of user facial video. By applying the training data, the pre-trained ViViT network is end-to-end supervised model training to obtain the final facial combination feature extraction model.

[0125] Based on the facial combination feature extraction model, facial combination feature extraction processing is performed on all the specified videos to obtain corresponding second feature vectors.

[0126] In this embodiment, the input of the facial combination feature extraction model is the X-second user facial video (specified video) determined by the sliding time window, and the facial combination feature extraction model extracts the combination features from the X-second user facial video. The output of the facial combination feature extraction model is the extraction result of the facial combination features, that is, an N-dimensional 0-1 vector. The N-dimensional is a total of N independent features related to mental health risk assessment, and each combination feature corresponds to a fixed position in the N-dimensional vector. A value of 0 indicates that the combination feature is recognized, and a value of 1 indicates that the combination feature is not recognized.

[0127] All the second feature vectors are concatenated based on the time sequence of the designated video to obtain a corresponding second feature matrix.

[0128] In this embodiment, after the facial combination feature extraction model is used to extract the facial combination features of all specified videos to obtain the second feature vector, the obtained second feature vector can be spliced ​​according to the time sequence of the sliding time window of the facial video to obtain a corresponding second feature matrix, wherein the rows of the second feature matrix correspond to each combination feature, and the columns are each sliding time window of the facial video arranged in time sequence.

[0129] The second feature matrix is ​​encoded based on a preset second encoding model to obtain the facial combination feature encoding vector.

[0130] In this embodiment, the second encoding model may specifically adopt a convolutional neural network (CNN). By using the convolutional neural network to encode the extracted second feature matrix, the required facial combination feature encoding vector may be obtained.

[0131] The present application splits the facial video into corresponding multiple designated videos based on a preset sliding time window; then calls a preset facial combination feature extraction model; and extracts facial combination features from all the designated videos based on the facial combination feature extraction model to obtain a corresponding second feature vector; then splices all the second feature vectors based on the time sequence of the designated videos to obtain a corresponding second feature matrix; and subsequently encodes the second feature matrix based on a preset second encoding model to obtain the facial combination feature encoding vector. The present application splits the facial video into corresponding multiple designated videos based on a preset sliding time window; then extracts facial combination features from all the designated videos based on the call of a facial combination feature extraction model to obtain a second feature vector; then splices all the second feature vectors based on the time sequence of the designated videos to obtain a second feature matrix; and then encodes the second feature matrix based on the use of the second encoding model, thereby achieving rapid and accurate completion of facial combination feature extraction for facial videos, improving the processing efficiency of facial combination feature extraction, and ensuring the data accuracy of the obtained facial combination feature encoding vector.

[0132] In some optional implementations, step S205 includes the following steps:

[0133] Get the preset vector concatenation strategy.

[0134] In this embodiment, the strategy content of the above-mentioned splicing strategy may include: splicing the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector in a target order. There is no specific limitation on the determination of the above-mentioned target order, and it can be set according to actual business needs, such as the facial overall feature vector-facial independent feature coding vector-facial combined feature coding vector.

[0135] Based on the vector splicing strategy, vector splicing processing is performed on the overall facial feature vector, the independent facial feature coding vector and the combined facial feature coding vector to obtain a corresponding splicing vector.

[0136] In this embodiment, according to the strategy content of the above-mentioned splicing strategy, the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector may be subjected to vector splicing processing to obtain a corresponding splicing vector.

[0137] The concatenated vector is used as the facial feature vector.

[0138] In this embodiment, the concatenated feature vector will contain information of all original feature vectors, forming a more comprehensive feature representation.

[0139] The present application obtains a preset vector splicing strategy; then based on the vector splicing strategy, the facial overall feature vector, the facial independent feature coding vector, and the facial combined feature coding vector are subjected to vector splicing processing to obtain a corresponding splicing vector; and the splicing vector is subsequently used as the facial feature vector. The present application performs vector splicing processing on the facial overall feature vector, the facial independent feature coding vector, and the facial combined feature coding vector based on the use of a vector splicing strategy, thereby achieving rapid and accurate generation of the required facial feature vector, improving the generation efficiency of the facial feature vector, and ensuring the accuracy of the obtained facial feature vector.

[0140] In some optional implementations of this embodiment, before step S206, the electronic device may further perform the following steps:

[0141] Get a sample set of pre-built risk assessments.

[0142] In this embodiment, a corresponding risk assessment sample set can be constructed by collecting a large amount of existing mental health risk data. The risk assessment sample set includes fused feature vectors (including facial overall feature vectors, facial independent feature matrices, and facial combined feature matrices obtained through three-level feature extraction networks of facial videos) of users with known mental health conditions of no risk, users with low risk, users with medium risk, and users with high risk of mental health, and their corresponding labels (no risk, low risk, medium risk, and high risk).

[0143] The risk assessment sample set is divided into a training set and a validation set based on a preset division ratio.

[0144] In this embodiment, there is no specific limitation on the numerical value selection of the above division ratio, which can be set according to actual business needs, for example, it can be set to 8:2.

[0145] Call the preset deep learning model.

[0146] In this embodiment, the deep learning model may specifically adopt a deep learning network DNN. The deep learning model includes an input layer, a hidden layer and an output layer, and the input layer can receive the concatenated feature vectors, and the output layer can output the risk label of the user's mental health status.

[0147] The deep learning model is trained based on the training set to obtain a trained first deep learning model.

[0148] In this embodiment, the deep learning model is trained by using the training set, the input is the fused feature vector, and the output is the risk label of the mental health condition. During the training process of the deep learning model, the prediction error is minimized by adjusting the model parameters (such as weights and biases). Cross-validation, early stopping and other strategies can be used to prevent overfitting, thereby obtaining a trained first deep learning model.

[0149] Based on the validation set, a performance evaluation is performed on the first deep learning model to obtain a corresponding performance evaluation result.

[0150] In this embodiment, the performance of the trained first deep learning model is evaluated by using the validation set, and indicators such as accuracy, recall rate, and F1 score are calculated as corresponding performance evaluation results.

[0151] The first deep learning model is optimized based on the performance evaluation result to obtain a second deep learning model that meets preset performance expectations.

[0152] In this embodiment, the first deep learning model can be tuned according to the obtained performance evaluation results to improve its generalization ability, thereby obtaining a second deep learning model that meets the preset performance expectations. Among them, the above-mentioned meeting the performance expectations can mean that the accuracy, recall rate, F1 score and other indicators of the model are all greater than the preset indicator threshold. There is no specific limitation on the value of the indicator threshold, which can be set according to actual business needs.

[0153] The second deep learning model is used as the risk assessment model.

[0154] In this embodiment, the risk assessment model generated by training can effectively process and analyze complex fusion feature data, effectively improving the accuracy and sensitivity of the risk assessment processing for the user's mental health status.

[0155] This application obtains a pre-constructed risk assessment sample set; and divides the risk assessment sample set into a training set and a validation set based on a preset division ratio; then calls a preset deep learning model; and trains the deep learning model based on the training set to obtain a trained first deep learning model; then performs a performance evaluation on the first deep learning model based on the validation set to obtain a corresponding performance evaluation result; further optimizes the first deep learning model based on the performance evaluation result to obtain a second deep learning model that meets the preset performance expectations; and finally uses the second deep learning model as the risk assessment model. This application divides a pre-constructed risk assessment sample set into a training set and a validation set by using a division ratio, and then trains the deep learning model based on the use of the training set to obtain a trained first deep learning model, and performs a performance evaluation on the first deep learning model based on the validation set to obtain a corresponding performance evaluation result, and then optimizes the first deep learning model according to the obtained performance evaluation result, so that a risk assessment model that meets performance expectations can be quickly and accurately constructed, which effectively improves the construction efficiency of the risk assessment model and ensures the model processing effect of the obtained risk assessment model.

[0156] In some optional implementations of this embodiment, after step S206, the electronic device may further perform the following steps:

[0157] Determine whether the risk assessment result meets the preset risk conditions.

[0158] In this embodiment, the above risk condition refers to the risk assessment result being of medium risk or above. If the content of the risk assessment result is no risk or low risk, it is determined that the risk assessment result does not meet the above risk condition. If the content of the risk assessment result is medium risk or high risk, it is determined that the risk assessment result meets the above risk condition.

[0159] If so, corresponding risk warning information is generated based on the risk assessment result.

[0160] In this embodiment, if the risk assessment result meets the risk condition, the risk assessment result can be filled into the preset risk warning information template to generate corresponding risk warning information. The template content of the risk warning information template can be written and generated according to the actual risk warning requirements.

[0161] Obtain the information push method corresponding to the user.

[0162] In this embodiment, the above-mentioned information push method may include SMS push, interface display, email push and the like.

[0163] Based on the information push method, the risk warning information is sent to the user.

[0164] In this embodiment, the risk warning information can be sent to the user by using the above-mentioned information push method.

[0165] This application determines whether the risk assessment result meets the preset risk conditions; if so, generates corresponding risk warning information based on the risk assessment result; then obtains the information push method corresponding to the user; and subsequently sends the risk warning information to the user based on the information push method. After the risk assessment result corresponding to the user is obtained by performing risk assessment processing on the facial feature vector based on the risk assessment model, if it is detected that the risk assessment result meets the preset risk conditions, the corresponding risk warning information will be automatically generated based on the risk assessment result, and the risk warning information will be intelligently sent to the user based on the acquired information push method, so that the user can timely understand the risk situation of his or her own mental health by consulting the received risk warning information and can take corresponding treatment measures, which effectively improves the user's experience.

[0166] In some optional implementations, the user information obtained is subject to the user's consent and complies with relevant laws and policies.

[0167] In addition, the non-Company software tools or components that appear in the embodiments of the present application are merely examples and do not represent actual use.

[0168] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0169] In addition, the intelligent mental health risk assessment system based on multi-level facial features proposed in this application has significant benefits: First, the system can comprehensively capture the changes in the user's facial expressions by analyzing the overall features, independent features and combined features in the facial video. This multi-level feature representation helps the system to deeply understand the subtle differences in facial expressions, thereby improving the accuracy of risk assessment; secondly, by applying a deep learning network, the system can effectively process complex facial feature data and achieve efficient and accurate mental illness risk assessment, and is also advanced and innovative in model technology; in addition, compared with traditional mental illness assessment methods, this non-invasive method of obtaining user data through facial videos reduces the user's psychological burden and makes more people willing to participate in mental health assessments, which not only helps to increase the early detection rate of mental illness, but also helps to promote the development of mental health management.

[0170] It should be emphasized that in order to further ensure the privacy and security of the above risk assessment results, the above risk assessment results can also be stored in a blockchain node.

[0171] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0172] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0173] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0174] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0175] Further references Figure 3 , as a response to the above Figure 2 In order to realize the method shown in the figure, the present application provides an embodiment of a risk assessment device, and the device embodiment is Figure 2Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0176] like Figure 3 As shown, the risk assessment device 300 described in this embodiment includes: a first acquisition module 301, a first extraction module 302, a second extraction module 303, a third extraction module 304, a fusion module 305 and an assessment module 306. Among them:

[0177] The first acquisition module 301 is used to acquire the user's facial video;

[0178] The first extraction module 302 is used to extract the overall facial features of the facial video to obtain a corresponding overall facial feature vector;

[0179] The second extraction module 303 is used to perform facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector;

[0180] The third extraction module 304 is used to perform facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector;

[0181] A fusion module 305 is used to fuse the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector to obtain a corresponding facial feature vector;

[0182] The evaluation module 306 is used to perform risk evaluation processing on the facial feature vector based on a preset risk evaluation model to obtain a risk evaluation result corresponding to the user.

[0183] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0184] In some optional implementations of this embodiment, the first extraction module 302 includes:

[0185] A preprocessing submodule, used for preprocessing the facial video to obtain a corresponding designated facial video;

[0186] The first calling submodule is used to call a preset facial overall feature extraction model;

[0187] A first encoding submodule is used to perform vector encoding processing on the specified facial video based on the facial overall feature extraction model to obtain a corresponding vector encoding result;

[0188] The first determination submodule is used to use the vector encoding result as the facial overall feature vector.

[0189] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0190] In some optional implementations of this embodiment, the second extraction module 303 includes:

[0191] A first splitting submodule, used for splitting the facial video into a corresponding plurality of video frames;

[0192] The second calling submodule is used to call a preset facial independent feature extraction model;

[0193] A first extraction submodule, configured to extract independent facial features from all the video frames based on the independent facial feature extraction model to obtain corresponding first feature vectors;

[0194] A first splicing submodule, configured to splice all the first feature vectors based on the time sequence of the video frames to obtain a corresponding first feature matrix;

[0195] The second encoding submodule is used to encode the first feature matrix based on a preset first encoding model to obtain the facial independent feature encoding vector.

[0196] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0197] In some optional implementations of this embodiment, the third extraction module 304 includes:

[0198] A second splitting submodule, used for splitting the facial video into a corresponding plurality of designated videos based on a preset sliding time window;

[0199] The third calling submodule is used to call a preset facial combination feature extraction model;

[0200] A second extraction submodule is used to extract facial combination features from all the specified videos based on the facial combination feature extraction model to obtain corresponding second feature vectors;

[0201] A second splicing submodule, configured to splice all the second feature vectors based on the time sequence of the designated video to obtain a corresponding second feature matrix;

[0202] The third encoding submodule is used to encode the second feature matrix based on a preset second encoding model to obtain the facial combination feature encoding vector.

[0203] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0204] In some optional implementations of this embodiment, the fusion module 305 includes:

[0205] The acquisition submodule is used to obtain the preset vector splicing strategy;

[0206] A processing submodule, configured to perform vector splicing processing on the facial overall feature vector, the facial independent feature coding vector, and the facial combined feature coding vector based on the vector splicing strategy to obtain a corresponding splicing vector;

[0207] The second determining submodule is configured to use the splicing vector as the facial feature vector.

[0208] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0209] In some optional implementations of this embodiment, the risk assessment device further includes:

[0210] A second acquisition module is used to acquire a pre-built risk assessment sample set;

[0211] A division module, used for dividing the risk assessment sample set into a training set and a validation set based on a preset division ratio;

[0212] The calling module is used to call the preset deep learning model;

[0213] A training module, used to train the deep learning model based on the training set to obtain a trained first deep learning model;

[0214] An evaluation module, configured to perform a performance evaluation on the first deep learning model based on the validation set to obtain a corresponding performance evaluation result;

[0215] An optimization module, configured to optimize the first deep learning model based on the performance evaluation result to obtain a second deep learning model that meets preset performance expectations;

[0216] A determination module is used to use the second deep learning model as the risk assessment model.

[0217] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0218] In some optional implementations of this embodiment, the risk assessment device further includes:

[0219] A judgment module is used to judge whether the risk assessment result meets the preset risk conditions;

[0220] A generating module, configured to generate corresponding risk warning information based on the risk assessment result;

[0221] A third acquisition module is used to acquire the information push mode corresponding to the user;

[0222] A sending module is used to send the risk warning information to the user based on the information push method.

[0223] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the risk assessment method in the aforementioned implementation mode, and are not described in detail here.

[0224] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0225] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0226] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0227] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the risk assessment method, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0228] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions or process data stored in the memory 41, such as computer-readable instructions for running the risk assessment method.

[0229] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0230] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the risk assessment method as described above.

[0231] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0232] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A risk assessment method, characterized in that: The steps include: Get the user's facial video; Performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector; Performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector; Performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector; The facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector are fused to obtain a corresponding facial feature vector; The facial feature vector is subjected to risk assessment processing based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

2. The risk assessment method according to claim 1, characterized in that: The step of performing facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector specifically includes: Preprocessing the facial video to obtain a corresponding designated facial video; Call the preset facial overall feature extraction model; Performing vector coding processing on the designated facial video based on the facial overall feature extraction model to obtain a corresponding vector coding result; The vector encoding result is used as the overall facial feature vector.

3. The risk assessment method according to claim 1, characterized in that: The step of performing facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector specifically includes: Splitting the facial video into a corresponding plurality of video frames; Call the preset facial independent feature extraction model; Performing facial independent feature extraction processing on all the video frames based on the facial independent feature extraction model to obtain corresponding first feature vectors; Based on the time sequence of the video frames, all the first feature vectors are concatenated to obtain a corresponding first feature matrix; The first feature matrix is ​​encoded based on a preset first encoding model to obtain the facial independent feature encoding vector.

4. The risk assessment method according to claim 1, characterized in that: The step of performing facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector specifically includes: Splitting the facial video into a corresponding plurality of designated videos based on a preset sliding time window; Call the preset facial combination feature extraction model; Performing facial combination feature extraction processing on all the specified videos based on the facial combination feature extraction model to obtain corresponding second feature vectors; Based on the time sequence of the specified video, all the second feature vectors are concatenated to obtain a corresponding second feature matrix; The second feature matrix is ​​encoded based on a preset second encoding model to obtain the facial combination feature encoding vector.

5. The risk assessment method according to claim 1, characterized in that: The step of fusing the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector to obtain a corresponding facial feature vector specifically includes: Get the preset vector splicing strategy; Based on the vector splicing strategy, vector splicing processing is performed on the facial overall feature vector, the facial independent feature coding vector, and the facial combined feature coding vector to obtain a corresponding splicing vector; The concatenated vector is used as the facial feature vector.

6. The risk assessment method according to claim 1, characterized in that: Before the step of performing risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user, the step further includes: Get access to a sample set of pre-built risk assessments; Dividing the risk assessment sample set into a training set and a validation set based on a preset division ratio; Call the preset deep learning model; Training the deep learning model based on the training set to obtain a trained first deep learning model; Performing a performance evaluation on the first deep learning model based on the validation set to obtain a corresponding performance evaluation result; Optimizing the first deep learning model based on the performance evaluation result to obtain a second deep learning model that meets preset performance expectations; The second deep learning model is used as the risk assessment model.

7. The risk assessment method according to claim 1, characterized in that: After the step of performing risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user, the method further includes: Determine whether the risk assessment results meet the preset risk conditions; If so, generating corresponding risk warning information based on the risk assessment result; Obtaining an information push method corresponding to the user; Based on the information push method, the risk warning information is sent to the user.

8. A risk assessment device, characterized in that: include: A first acquisition module is used to acquire a user's facial video; A first extraction module is used to perform facial overall feature extraction processing on the facial video to obtain a corresponding facial overall feature vector; A second extraction module is used to perform facial independent feature extraction processing on the facial video to obtain a corresponding facial independent feature coding vector; A third extraction module is used to perform facial combination feature extraction processing on the facial video to obtain a corresponding facial combination feature coding vector; A fusion module, used for fusing the facial overall feature vector, the facial independent feature coding vector and the facial combined feature coding vector to obtain a corresponding facial feature vector; An assessment module is used to perform risk assessment processing on the facial feature vector based on a preset risk assessment model to obtain a risk assessment result corresponding to the user.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the risk assessment method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the risk assessment method according to any one of claims 1 to 7 are implemented.

Images