An interactive risk assessment method based on human behavior and scenario analysis

Abstract

The embodiment of the present application provides an interactive risk assessment method for human behavior and scene analysis, which relates to the field of risk assessment technology. The method includes converting the answering voice of the testee into an answering text; determining the text matching degree between the answering text and the answering reference text; separating the character image and the background image in the on-site test video of the testee to obtain the target character image and the target background image; matching the feature points in the target background image with the feature points in the blacklist background image to obtain the feature point matching degree; using the target character image as the input of a pre-trained human behavior recognition model to perform operations to obtain the behavior feature parameters corresponding to the target character image; and determining the fraud risk parameters corresponding to the testee based on the text matching degree, the feature point matching degree and the behavior feature parameters. The method provided by the embodiment of the present application can efficiently and accurately perform fraud risk assessment.

Description

Technical Field

[0001] This document relates to the field of risk assessment technology, and in particular to an interactive risk assessment method for human behavior and scenario analysis. Background Art

[0002] In order to ensure the authenticity of the information provided, it is often necessary to conduct a fraud risk assessment on the information provided by the user to assess whether there is a risk of fraud, such as the authenticity assessment of personal certification materials during the credit process.

[0003] At present, a commonly used method for assessing fraud risk is risk assessment based on a scale. This assessment method has low efficiency and accuracy because the scale scheme takes a long time and the scale assessment method is too subjective and easy to be influenced by personal bias.

[0004] Therefore, how to provide an effective solution to improve the efficiency and accuracy of fraud risk assessment has become a problem to be solved urgently in the prior art. Summary of the invention

[0005] The embodiment of the present application provides an interactive risk assessment method of human behavior and scene analysis to solve the problems of low efficiency and accuracy of fraud risk assessment in the prior art.

[0006] To solve the above technical problems, the embodiments of the present application are implemented as follows:

[0007] The present application embodiment provides an interactive risk assessment method for human behavior and scene analysis, including:

[0008] Converting the testee's voice response to a designated question into a response text, wherein the designated question is generated based on the background information provided by the testee;

[0009] Determining a text matching degree between the reply text and the reply reference text based on a Euclidean distance between the reply text and a reply reference text corresponding to the specified question;

[0010] Separating the person image and the background image in the on-site test video of the test subject by a semantic segmentation model to obtain a target person image and a target background image;

[0011] Extracting feature points in the target background image by using a scale-invariant feature transformation algorithm;

[0012] Matching the feature points in the target background image with the feature points in the preset blacklist background image to obtain a matching degree of the feature points;

[0013] The target person image is used as an input of a pre-trained human behavior recognition model to perform operations to obtain behavior feature parameters corresponding to the target person image;

[0014] A fraud risk parameter corresponding to the testee is determined based on the text matching degree, the feature point matching degree and the behavior feature parameter, and the fraud risk parameter represents the probability that the testee has fraudulent behavior.

[0015] Optionally, the method further comprises:

[0016] Generating the designated question and the answer reference text corresponding to the designated question according to the background information;

[0017] converting the specified question into an audio file;

[0018] The audio file is played so that the test subject can answer the questions in the audio file to obtain the answering voice.

[0019] Optionally, determining the fraud risk parameter corresponding to the testee based on the text matching degree, the feature point matching degree and the behavior feature parameter includes:

[0020] Performing data fusion on the text matching degree, the feature point matching degree and the behavior feature parameter to obtain fused data;

[0021] The fused data is classified by a support vector machine algorithm to obtain the fraud risk parameter.

[0022] Optionally, fusing the text matching degree, the feature point matching degree and the behavior feature parameter to obtain fused data includes:

[0023] Fusing multiple sets of feature data corresponding to different time periods with the text matching degree to obtain multiple fused data;

[0024] The method of classifying the fused data by a support vector machine algorithm to obtain the fraud risk parameter includes:

[0025] Classifying the plurality of fused data by a support vector machine algorithm to obtain the fraud risk parameter;

[0026] A set of feature data includes feature point matching degrees and behavior feature parameters corresponding to the same time period.

[0027] Optionally, determining the fraud risk parameter corresponding to the testee based on the text matching degree, the feature point matching degree and the behavior feature parameter includes:

[0028] Performing data fusion on the text matching degree, the feature point matching degree and the behavior feature parameter to obtain fused data;

[0029] The fused data is classified by an extreme gradient boosting algorithm to obtain the fraud risk parameter.

[0030] Optionally, the step of using the target person image as an input of a pre-trained human behavior recognition model to perform operation to obtain behavior feature parameters corresponding to the target person image includes:

[0031] The character image is used as an input of a pre-trained action recognition model or a long short-term memory network model to perform operations to obtain the behavior feature parameters.

[0032] The above technical solution adopted in the embodiment of the present application can achieve the following beneficial effects:

[0033] By determining the text matching degree between the test subject's reply text and the reply reference text, the character image and background image in the field test video are separated to obtain the target character image and the target background image, and the feature point matching degree between the target background image in the field test video and the blacklist background image is determined, and the target character image is used as the input of the pre-trained human behavior recognition model for operation to obtain the behavior feature parameters corresponding to the target character image, and then the fraud risk parameters corresponding to the test subject are determined based on the obtained text matching degree, feature point matching degree and behavior feature parameters. In this process, risk assessment is carried out by comprehensively considering parameters from different dimensions, so that the fraud risk can be accurately assessed, and the entire assessment process is convenient and fast. At the same time, personal subjective bias will not be incorporated into the assessment process, further ensuring the accuracy of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings described herein are used to provide a further understanding of this document and constitute a part of this document. The exemplary embodiments and descriptions of this document are used to explain this document and do not constitute an improper limitation on this document. In the drawings:

[0035] Figure 1 A flowchart of an interactive risk assessment method for human behavior and scene analysis provided in accordance with an embodiment of the present application.

[0036] Figure 2 A schematic diagram of the structure of an electronic device provided for one embodiment of the present application.

[0037] Figure 3 A schematic diagram of the structure of an interactive risk assessment device for human behavior and scene analysis provided in one embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of this document clearer, the technical solutions of this document will be clearly and completely described below in combination with the specific embodiments of this document and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this document, not all of them. Based on the embodiments in this document, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this document.

[0039] In order to improve the efficiency and accuracy of fraud risk assessment, an embodiment of the present application provides an interactive risk assessment method of human behavior and scene analysis, which can accurately and efficiently perform fraud risk assessment.

[0040] The interactive risk assessment method for human behavior and scene analysis provided in an embodiment of the present application will be described in detail below.

[0041] The interactive risk assessment method for human behavior and scene analysis provided in the embodiments of the present application can be applied to a user terminal, which may be, but is not limited to, a personal computer, a smart phone, a tablet computer, a laptop computer, a personal digital assistant, etc.

[0042] It can be understood that the execution entity does not constitute a limitation on the embodiments of the present application.

[0043] Optionally, the process of the interactive risk assessment method for human behavior and scene analysis is as follows: Figure 1 As shown, the following steps may be included:

[0044] Step S101: converting the test subject's voice response to a designated question into a response text.

[0045] The designated questions are generated based on the background information provided by the test subject. For example, if the background information provided by the test subject includes home address, work unit, previous work unit, etc., the designated questions may be "Where is your home address?", "Where is your work unit address?", etc.

[0046] The user terminal has an audio playback and acquisition function. After generating a specified question, the specified question in text format can be converted into an audio file and played, so that the testee can answer the question in the audio file. When the testee answers, the user terminal obtains the testee's answer voice, and then converts the testee's answer voice to the specified question into an answer text.

[0047] Step S102, based on the Euclidean distance between the reply text and the reply reference text corresponding to the specified question, determine the text matching degree between the reply text and the reply reference text.

[0048] The user terminal also generates and specifies question answer reference text based on the information recorded in the background data.

[0049] If there is false content in the background information provided by the test subject, the test subject may not be able to accurately remember the false content, so there will be a large difference between the content of the response text replaced when answering the specified question and the content of the response reference text. Therefore, the text matching degree between the response text and the response reference text corresponding to the specified question can be determined by calculating the Euclidean distance between the response text and the response reference text corresponding to the specified question, so as to facilitate subsequent fraud detection.

[0050] Among them, the larger the Euclidean distance between the reply text and the reply reference text, the greater the difference between the two, and thus the lower the text matching degree between the reply text and the reply reference text.

[0051] In the embodiment of the present application, there can be multiple reply texts and reply reference texts, and they correspond one to one. Therefore, the text matching degree between the obtained reply text and the reply reference text can also be multiple.

[0052] Step S103, separating the person image and the background image in the on-site test video of the testee by using a semantic segmentation model to obtain a target person image and a target background image.

[0053] Step S104: extracting feature points in the target background image by using a scale-invariant feature transformation algorithm.

[0054] In the embodiment of the present application, feature points in the target background image may be extracted by a scale-invariant feature transform (SIFT) algorithm at regular time intervals (eg, 3 seconds) to obtain feature points in the target background image in different time periods.

[0055] Step S105 , matching the feature points in the target background image with the feature points in the preset blacklist background image to obtain a matching degree of the feature points.

[0056] In the embodiment of the present application, a background image is extracted in advance from a plurality of test videos known to contain fraud to obtain a plurality of blacklist background images. After extracting feature points in the target background image, the feature points in the target background image can be matched with feature points in a plurality of blacklist background images to obtain a feature point matching degree.

[0057] The feature point matching degree may be a multi-dimensional vector, in which the value of each dimension represents the matching degree of the feature point in the target background image with the feature point in a blacklist background image. Assuming there are four blacklist background images, the feature point in the target background image is matched with the feature points in the four blacklist background images, and the feature point matching degree obtained may be a four-dimensional vector.

[0058] Step S106, using the target person image as an input of a pre-trained human behavior recognition model to perform operations to obtain behavior feature parameters corresponding to the target person image.

[0059] In the embodiment of the present application, a human behavior recognition model for human behavior recognition is pre-trained, and the human behavior may be, but is not limited to, scratching the ear, supporting the chin, communicating with irrelevant persons, scratching the head, answering hesitantly, and the like.

[0060] After obtaining the target person image, the user terminal can use the target person image as the input of the pre-trained human behavior recognition model for operation at a certain interval (the interval is synchronized with the time of extracting the feature points in the target background image) to obtain multiple behavior feature parameters corresponding to the target person image. The behavior feature parameter can be a multi-dimensional vector, and the value of each dimension in the multi-dimensional vector is used to indicate whether a certain behavior occurs. For example, the human behavior recognition model recognizes behaviors in five dimensions, such as scratching ears, supporting chins, communicating with irrelevant people, scratching heads, and hesitant answers. The obtained behavior feature parameters can be represented by a five-dimensional vector.

[0061] In an embodiment of the present application, the human behavior recognition model may be, but is not limited to, a temporal shift module (TSM) action recognition model or a long short-term memory network (LSTM) model.

[0062] Step S107, determining the fraud risk parameter corresponding to the test subject according to the text matching degree, the feature point matching degree and the behavior feature parameter.

[0063] Among them, the fraud risk parameter represents the probability that the test subject has committed fraudulent behavior.

[0064] Specifically, the text matching degree, feature point matching degree and behavior feature parameters may be fused to obtain fused data, and then the fused data may be classified to obtain the fraud risk parameters corresponding to the person being tested.

[0065] When performing fusion, the text matching degree can be quantified, and then the quantified text matching degree is vector-fused with the feature point matching degree and the behavior feature parameters to obtain fused data.

[0066] For example, the quantized text matching degree is (1, 0), the feature point matching degree is (0, 0, 1), and the behavior feature parameter is (1, 1, 1). Then the fused data obtained after vector fusion can be expressed as (1, 0, 0, 0, 1, 1, 1, 1).

[0067] Since a feature point matching degree and a behavior feature parameter can be determined at a certain time interval, when classifying, multiple sets of feature data corresponding to different time periods can be fused with the text matching degree to obtain multiple fused data, and then the multiple fused data are classified to obtain the fraud risk parameter. Among them, a set of feature data includes the feature point matching degree and behavior feature parameter corresponding to the same time period.

[0068] The fraud risk parameter represents the probability of the test subject committing fraud. Furthermore, the user terminal can also determine whether there is a fraud risk based on the probability of the fraud. If the probability of the test subject committing fraud is greater than or equal to a set threshold, it is considered that there is a fraud risk.

[0069] The user terminal can also classify multiple risk levels according to the probability of fraud. For example, when the probability of fraud is less than 2%, it can be classified as low risk, when the probability of fraud is 2%-5%, it can be classified as medium risk, and when the probability of fraud is greater than 5%, it can be classified as high risk. In this way, it can be widely used in various risk assessments as a basis for risk assessment.

[0070] The interactive risk assessment method of human behavior and scene analysis provided by the embodiment of the present application determines the text matching degree between the reply text of the testee and the reply reference text, separates the character image and the background image in the field test video to obtain the target character image and the target background image, and determines the feature point matching degree between the target background image and the blacklist background image in the field test video, and uses the target character image as the input of the pre-trained human behavior recognition model for operation, obtains the behavior feature parameters corresponding to the target character image, and then determines the fraud risk parameters corresponding to the testee based on the obtained text matching degree, feature point matching degree and behavior feature parameters. In this process, risk assessment is carried out by comprehensively considering parameters from different dimensions, so that the fraud risk can be accurately assessed, and the entire assessment process does not need to wait for a long time, which is convenient and fast, and has good timeliness. At the same time, personal subjective bias will not be incorporated into the assessment process, further ensuring the accuracy of the assessment. In addition, it is also possible to determine whether there is a fraud risk or divide multiple risk levels based on the probability of fraudulent behavior, so that it can be widely used in various risk assessments as a basis for risk assessment.

[0071] Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 2At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0072] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0073] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0074] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming an interactive risk assessment device for human behavior and scene analysis at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0075] Converting the testee's voice response to a designated question into a response text, wherein the designated question is generated based on the background information provided by the testee;

[0076] Determining a text matching degree between the reply text and the reply reference text based on a Euclidean distance between the reply text and a reply reference text corresponding to the specified question;

[0077] Separating the person image and the background image in the on-site test video of the test subject by a semantic segmentation model to obtain a target person image and a target background image;

[0078] Extracting feature points in the target background image by using a scale-invariant feature transformation algorithm;

[0079] Matching the feature points in the target background image with the feature points in the preset blacklist background image to obtain a matching degree of the feature points;

[0080] The target person image is used as an input of a pre-trained human behavior recognition model to perform operations to obtain behavior feature parameters corresponding to the target person image;

[0081] A fraud risk parameter corresponding to the testee is determined based on the text matching degree, the feature point matching degree and the behavior feature parameter, and the fraud risk parameter represents the probability that the testee has fraudulent behavior.

[0082] The above application Figure 2 The method performed by the interactive risk assessment device for human behavior and scene analysis disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in software form. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in one or more embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with one or more embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0083] The electronic device may also perform Figure 1 The method of the embodiment shown in the figure realizes an interactive risk assessment device for human behavior and scene analysis. Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0084] Of course, in addition to software implementation methods, the electronic device of the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0085] The embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by a portable electronic device including a plurality of application programs, enable the portable electronic device to execute Figure 1 The method of the embodiment shown is specifically used to perform the following operations:

[0086] Converting the testee's voice response to a designated question into a response text, wherein the designated question is generated based on the background information provided by the testee;

[0087] Determining a text matching degree between the reply text and the reply reference text based on a Euclidean distance between the reply text and a reply reference text corresponding to the specified question;

[0088] Separating the person image and the background image in the on-site test video of the test subject by a semantic segmentation model to obtain a target person image and a target background image;

[0089] Extracting feature points in the target background image by using a scale-invariant feature transformation algorithm;

[0090] Matching the feature points in the target background image with the feature points in the preset blacklist background image to obtain a matching degree of the feature points;

[0091] The target person image is used as an input of a pre-trained human behavior recognition model to perform operations to obtain behavior feature parameters corresponding to the target person image;

[0092] A fraud risk parameter corresponding to the testee is determined based on the text matching degree, the feature point matching degree and the behavior feature parameter, and the fraud risk parameter represents the probability that the testee has fraudulent behavior.

[0093] Figure 3 This is a schematic diagram of the structure of an interactive risk assessment device for human behavior and scene analysis provided by an embodiment of the present application. Figure 3 In one software implementation, the interactive risk assessment device for human behavior and scene analysis provided may include:

[0094] A text conversion module, used for converting the testee's answer voice to a specified question into a text answer, wherein the specified question is generated based on the background information provided by the testee;

[0095] A first determination module is used to determine the text matching degree between the reply text and the reply reference text based on the Euclidean distance between the reply text and the reply reference text corresponding to the specified question;

[0096] A separation module, used for separating the person image and the background image in the on-site test video of the testee through a semantic segmentation model to obtain a target person image and a target background image;

[0097] An extraction module, used for extracting feature points in the target background image by using a scale-invariant feature transformation algorithm;

[0098] A matching module, used for matching the feature points in the target background image with the feature points in the preset blacklist background image to obtain a matching degree of the feature points;

[0099] A calculation module, used to use the target person image as an input of a pre-trained human behavior recognition model to perform calculations to obtain behavior feature parameters corresponding to the target person image;

[0100] The second determination module is used to determine a fraud risk parameter corresponding to the testee based on the text matching degree, the feature point matching degree and the behavior feature parameter, wherein the fraud risk parameter represents the probability that the testee has fraudulent behavior.

[0101] Optional, interactive risk assessment device for human behavior and scenario analysis also includes:

[0102] A generating module, used for generating the designated question and the answer reference text corresponding to the designated question according to the background information;

[0103] An audio conversion module, used for converting the specified question into an audio file;

[0104] The playing module is used to play the audio file so that the test subject can answer the questions in the audio file to obtain the answering voice.

[0105] The technical solution provided by the embodiment of the present application determines the text matching degree between the reply text of the testee and the reply reference text, separates the character image and the background image in the field test video to obtain the target character image and the target background image, and determines the feature point matching degree between the target background image in the field test video and the blacklist background image, and uses the target character image as the input of the pre-trained human behavior recognition model for operation, obtains the behavior feature parameters corresponding to the target character image, and then determines the fraud risk parameters corresponding to the testee based on the obtained text matching degree, feature point matching degree and behavior feature parameters. In this process, risk assessment is carried out by comprehensively considering parameters from different dimensions, so that the fraud risk can be accurately assessed, and the entire assessment process does not need to wait for a long time, which is convenient and fast, and has good timeliness. At the same time, personal subjective bias will not be incorporated into the assessment process, further ensuring the accuracy of the assessment. In addition, it is also possible to determine whether there is a fraud risk or divide multiple risk levels based on the probability of fraudulent behavior, so that it can be widely used in various risk assessments as a basis for risk assessment.

[0106] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0107] In short, the above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0108] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0109] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0110] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0111] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. An interactive risk assessment method for human behavior and scene analysis, characterized in that: include: Converting the testee's voice response to a designated question into a response text, wherein the designated question is generated based on the background information provided by the testee; Determining a text matching degree between the reply text and the reply reference text based on a Euclidean distance between the reply text and a reply reference text corresponding to the specified question; Separating the person image and the background image in the on-site test video of the test subject by a semantic segmentation model to obtain a target person image and a target background image; Extracting feature points in the target background image by using a scale-invariant feature transformation algorithm; Matching the feature points in the target background image with the feature points in the preset blacklist background image to obtain a matching degree of the feature points; The target person image is used as an input of a pre-trained human behavior recognition model to perform operations to obtain behavior feature parameters corresponding to the target person image; Determine a fraud risk parameter corresponding to the testee according to the text matching degree, the feature point matching degree and the behavior feature parameter, wherein the fraud risk parameter represents the probability that the testee has a fraudulent behavior; The step of determining the fraud risk parameter corresponding to the test subject based on the text matching degree, the feature point matching degree and the behavior feature parameter includes: Performing data fusion on the text matching degree, the feature point matching degree and the behavior feature parameter to obtain fused data; The fused data is classified by a support vector machine algorithm to obtain the fraud risk parameter.

2. The method according to claim 1, characterized in that The method further comprises: Generating the designated question and the answer reference text corresponding to the designated question according to the background information; converting the specified question into an audio file; The audio file is played so that the test subject can answer the questions in the audio file to obtain the answering voice.

3. The method according to claim 1, characterized in that: The step of fusing the text matching degree, the feature point matching degree and the behavior feature parameter to obtain fused data includes: Fusing multiple sets of feature data corresponding to different time periods with the text matching degree to obtain multiple fused data; The method of classifying the fused data by a support vector machine algorithm to obtain the fraud risk parameter includes: Classifying the plurality of fused data by a support vector machine algorithm to obtain the fraud risk parameter; A set of feature data includes feature point matching degrees and behavior feature parameters corresponding to the same time period.

4. The method according to claim 1, characterized in that: The step of determining the fraud risk parameter corresponding to the test subject based on the text matching degree, the feature point matching degree and the behavior feature parameter includes: Performing data fusion on the text matching degree, the feature point matching degree and the behavior feature parameter to obtain fused data; The fused data is classified by an extreme gradient boosting algorithm to obtain the fraud risk parameter.

5. The method according to claim 1, characterized in that The step of using the target person image as an input of a pre-trained human behavior recognition model to perform operations to obtain behavior feature parameters corresponding to the target person image includes: The character image is used as an input of a pre-trained action recognition model or a long short-term memory network model to perform operations to obtain the behavior feature parameters.

Images