A call center-based video ivr risk identification system

Abstract

The application discloses a kind of video IVR risk identification systems based on call center, it is related to risk identification technical field, the application includes data acquisition module, data analysis module, risk identification module, early warning terminal and database, microphone array is installed in suitable position, the voice of user is collected by microphone array, and each voice instruction of user is identified, while obtaining the real-time operation data and real-time voice of each user in each time operation process in history from database, and the age of each user in history, construct voice-operation relationship diagram, when user operates, whether there is the risk of identity theft is detected, if there is, early warning is carried out and operation page is closed, if there is not, whether user is abnormal is judged, if abnormal, early warning is carried out and operation page is closed, the accuracy of abnormal operation detection and the effectiveness of operation action timeout judgment are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of risk identification technology, and specifically to a video IVR risk identification system based on a call center. Background Technology

[0002] With the development of communication technology, the service forms of call centers are constantly being upgraded, and video IVR systems have emerged. They provide users with more intuitive and richer self-service through video interaction. However, while video IVR brings convenience, it also faces many risks, such as identity fraud, information leakage and abnormal operation. Therefore, there is an urgent need for a video IVR risk identification system.

[0003] Traditional call center-based video IVR risk identification systems use facial detection technology to determine if there is a risk of identity theft when a user logs in. If so, an alert is issued. If not, the system analyzes the user's voice and actions independently to determine if the user's actions are abnormal and issues an alert if abnormality is detected. Clearly, this call center-based video IVR risk identification system has at least the following shortcomings: 1. Traditional call center-based video IVR risk identification systems analyze voice and actions independently when detecting risks, lacking analysis of the correlation between the two. Therefore, when determining whether user actions are abnormal, the accuracy of abnormality detection cannot be guaranteed.

[0004] 2. Traditional call center-based video IVR risk identification systems keep the operation time of each action constant when detecting whether a user's operation has timed out. However, users of different ages require different operation times for each action, so the effectiveness of the operation timeout judgment cannot be guaranteed, and consequently the accuracy of operation anomaly detection cannot be guaranteed. Summary of the Invention

[0005] To address the aforementioned technical shortcomings, the present invention aims to provide a video IVR risk identification system based on a call center.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a video IVR risk identification system based on a call center, including the following modules: a data acquisition module, a data analysis module, a risk identification module, an early warning terminal, and a database.

[0007] The data acquisition module is used to record the user's operation data in real time when the user performs an operation, and to select a suitable location to install a microphone array at that location to collect the user's voice in real time through the microphone array.

[0008] The data analysis module is used to obtain the operation actions and voice commands of each user in each historical operation process from the database, and to construct a voice-operation relationship diagram based on the operation actions and voice commands of each user in each historical operation process.

[0009] The risk identification module is used to obtain the personal information entered by the user and the user's real-time voice when the user performs an operation, and to determine whether there is a risk of identity theft. If there is, an alert is issued and the operation page is closed. If not, the module determines whether the user's operation is abnormal based on the user's real-time voice, real-time operation data and voice-operation relationship diagram. If the user's operation is abnormal, an alert is issued and the operation page is closed.

[0010] The warning terminal is used to issue a warning and close the operation page when the user is not the user or when the user's operation is abnormal.

[0011] The database is used to store the operation actions, the operation sequence of each operation action, the operation duration of each voice command and operation action, and the age of each historical user in each operation process retrieved from the database.

[0012] The beneficial effects of this invention are as follows: 1. This invention provides a video IVR risk identification system based on a call center. A microphone array is installed in a suitable location to collect the user's voice and identify the user's voice commands. At the same time, real-time operation data and real-time voice of each user in each historical operation process are obtained from the database, as well as the age of each user in history. A voice-operation relationship graph is constructed. When the user performs an operation, the system detects whether there is a risk of identity theft. If so, an alert is issued and the operation page is closed. If not, the system determines whether the user's operation is abnormal. If abnormal, an alert is issued and the operation page is closed. This ensures the accuracy of operation abnormality detection and the effectiveness of operation timeout judgment.

[0013] This invention refers to each historical user as a marked user, and retrieves each marked user's historical operation process, including each operation action, the order of operation actions, and each voice command from the database. It analyzes the relationship between each voice command and each operation action, and analyzes the operation duration of each operation action for users of different age groups. Based on the relationship between each voice command and the order of operation actions and the operation duration of each operation action for users of different age groups, a voice-operation relationship diagram is constructed. Based on the voice-operation relationship diagram, it is determined whether the user's operation is abnormal, thus ensuring the accuracy of operation anomaly detection.

[0014] This invention divides each marked user into user groups based on age. Within each user group, it retrieves the operation actions and operation durations of each marked user's historical operation process from the database. By analyzing the operation durations of each operation action for users of different age groups, the effectiveness of operation action timeout judgment is ensured. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 As shown, the present invention provides a video IVR risk identification system based on a call center, including: a data acquisition module, a data analysis module, a risk identification module, an early warning terminal, and a database.

[0019] The data acquisition module is connected to the data analysis module, the data analysis module is connected to the risk identification module, the risk identification module is connected to the early warning terminal, and the database is connected to the data analysis module.

[0020] The data acquisition module is used to record the user's operation data in real time when the user performs an operation, and to select a suitable location to install a microphone array at that location to collect the user's voice in real time through the microphone array.

[0021] In one specific embodiment, the process of selecting a suitable location is as follows: obtain the operation terminal page display, and set each mounting point on the border of the operation terminal page display according to a preset distance interval, and install a microphone array at each mounting point. When the user performs voice operation in the operation area, the microphone array at each mounting point collects the user's voice.

[0022] It should be noted that the preset distance threshold is a critical value used to determine whether the settings of each installation point are reasonable, and it is set by the relevant staff.

[0023] The signal strength and signal-to-noise ratio of the user's voice collected by the microphone array at each installation point are obtained and compared with the target signal strength and target signal-to-noise ratio. The installation points of each microphone array where the signal strength is greater than the target signal strength and the signal-to-noise ratio is greater than the target signal-to-noise ratio are identified and referred to as each marked installation point.

[0024] It should be noted that the sensitivity of the microphone array is obtained from the microphone array instruction manual, and the voltage amplitude of the voice signal received by the microphone array is measured using a multimeter. The sound pressure level is then obtained by dividing the voltage amplitude by the sensitivity. Sound pressure level represents signal strength, and the signal-to-noise ratio is obtained using MATLAB software.

[0025] It should also be noted that the target signal strength and target signal-to-noise ratio are threshold values ​​used to determine whether the sound collected by the microphone array meets the requirements, and are set by the staff.

[0026] Obtain the angle between each marker installation point and the user, and compare it with a preset angle threshold. Marker installation points whose angle with the user is less than the preset angle threshold are called backup installation points. Randomly select one backup installation point from the backup installation points. This backup installation point is a suitable installation position for the microphone array.

[0027] It should be noted that the user's center point is obtained, connected to the marker installation point, and a perpendicular line is drawn from the user's center point to the line containing the marker installation point, forming a right triangle. The angle between the user and the marker installation point is calculated in this right triangle, and this angle is a horizontal angle.

[0028] It should also be noted that the preset angle threshold is 30°, at which point the target speech is within the main lobe, resulting in the highest gain and the best noise suppression effect.

[0029] The data analysis module is used to obtain the operation actions and voice commands of each user in each historical operation process from the database, and to construct a voice-operation relationship diagram based on the operation actions and voice commands of each user in each historical operation process.

[0030] It should be noted that voice commands include opening a certain interface and obtaining certain information, and operation actions include signing, entering information, and clicking various function buttons, such as confirming, deleting, and returning.

[0031] In a specific embodiment, the data analysis module performs the following process: referring to each historical user as a marked user, and obtaining each operation action, the operation sequence of each operation action, and each voice command from the database during each historical operation of each marked user, and analyzing the relationship between each voice command and each operation action.

[0032] The ages of each tagged user are retrieved from the database and compared. Users whose ages are in the same age range are grouped into user groups. This method is used to obtain each user group.

[0033] It should be noted that the age groups are divided into 0-12 years old, 12-18 years old, 18-40 years old, 40-60 years old and over 60 years old, and the operation speed of people in different age groups is different.

[0034] Within each user group, and further divided according to age, each marked user is further divided into user groups. Within each user group, the operation actions and operation durations of each marked user's historical operation processes are retrieved from the database, and the operation durations of each operation action of users in different age groups are analyzed.

[0035] Based on the relationship between each voice command and the sequence of each operation action, and the operation time of each operation action for users of different age groups, a voice-operation relationship diagram is constructed.

[0036] It should be noted that the voice groups of each marking operation process are connected to the operation groups of each marking operation process, and the operation time of users of different age groups is marked in each operation action in each operation group.

[0037] The above-mentioned analysis of the relationship between each voice command and each operation action is specifically carried out as follows: Each historical operation process of each marked user is referred to as a marked operation process. Based on each voice command in each marked operation process, the similarity value between each marked operation process is obtained. Each marked operation process with a similarity value of 1 is divided into an operation process group. Each operation process group is obtained in this way.

[0038] In each operation process group, each operation action and the operation sequence of each operation action in each marked operation process are obtained and compared. Marked operation processes with the same operation actions and the same operation sequence are divided into a sub-operation process group. Sub-operation process groups of each operation process group are obtained in this way.

[0039] It should be noted that, assuming a certain operation process group contains three marked operation processes, and the actions and their order in the first marked operation process are: enter password, click OK, and click Back; the actions and their order in the second marked operation process are: enter password, click OK, and click Back; and the actions and their order in the third marked operation process are: click OK, enter password, and click Back, then the first marked operation process and the third marked operation process in the operation process group belong to one sub-operation process group. This example is for illustrative purposes only and is not the only possible approach.

[0040] Each voice command corresponding to each operation process group is divided into a voice group, and each operation action corresponding to each sub-operation process group of each operation process group and the order of each operation action are divided into each operation group. Then, there is a voice operation relationship between each voice command in the voice group of each operation process group and each operation action in each operation group of each operation group.

[0041] The process of obtaining the similarity value between each marking operation process is as follows: compare each voice command of each marking operation process. If the voice commands of multiple marking operation processes are the same and the order of the voice commands is also the same, then the similarity value between the multiple marking operation processes is 1. If the voice commands of multiple marking operation processes are different or the order of the voice commands is different, then the similarity value between the multiple marking operation processes is 0.

[0042] The above-mentioned analysis of the operation time of each operation action of users in different age groups is specifically carried out as follows: In each user group, each operation action in each marked operation process is obtained, and the same operation actions are divided into a marked group. Each marked group is obtained in this way, and each marked operation process in each marked group is obtained. The operation time of the operation action corresponding to each marked group in each marked operation process in each marked group is obtained and compared to obtain the maximum operation time of each operation action. The maximum operation time of each operation action in each user group is obtained by a method, and it is used as the operation time of each operation action of each user group corresponding to the age group. The operation time of each operation action of users in different age groups is obtained in this way.

[0043] The risk identification module is used to obtain the personal information entered by the user and the user's real-time voice when the user performs an operation, and to determine whether there is a risk of identity theft. If there is, an alert is issued and the operation page is closed. If not, the module determines whether the user's operation is abnormal based on the user's real-time voice, real-time operation data and voice-operation relationship diagram. If the user's operation is abnormal, an alert is issued and the operation page is closed.

[0044] In a specific embodiment, the risk identification module operates as follows: When a user logs in, it determines whether the user is logging in for the first time. If it is, face detection is performed during the login process to determine whether there is a risk of identity theft. If so, a warning is issued and the operation page is closed. If there is no risk of identity theft, the user is prompted to read aloud a set text, and the user's voice is collected to obtain the user's voiceprint features. The user's personal information and voiceprint features are recorded in the user-voiceprint feature table in the database.

[0045] It should be noted that during the login process, the personal information entered by the user is compared with the personal information already stored in the database. If the personal information entered by the user is not in the database, it means that the user is logging in for the first time. If the personal information entered by the user is in the database, it means that the user is not logging in for the first time. The personal information includes the user's ID number, name, and age, etc.

[0046] It should also be noted that the face detection process is as follows: The user places the front of their ID card in the recognition area and faces the camera. YOLO-Face, based on deep learning, acquires the facial features of the face on the ID card. Simultaneously, the system generates dynamic instructions in real time. The user performs corresponding actions according to the instructions, and the face detection algorithm tracks the movement trajectory of key facial points to verify the continuity, completeness, and physiological rationality of the actions. If the action is not completed or the trajectory is abnormal, the user is at risk of identity theft. If the action is completed and the trajectory is normal, the user's facial features in the camera are extracted and compared with the facial features of the face on the ID card. If the user's facial features in the camera are different from the facial features of the face on the ID card, the user is at risk of identity theft.

[0047] The process involves removing DC signals caused by circuit offset using a high-pass filter to avoid baseline drift, segmenting continuous speech into short-time frames, adding a Hanning window to each frame to reduce spectral leakage at frame edges, using a dual threshold method of short-time energy and zero-crossing rate to distinguish speech segments from silence segments, retaining only valid speech frames, employing spectral subtraction or Wiener filtering for low signal-to-noise ratio speech to subtract noise spectrum estimates in the frequency domain while preserving speech features, eliminating acoustic echoes between the microphone and speaker through adaptive filtering, normalizing the amplitude of the speech signal, mapping the speech spectrum to the Mel scale, calculating cepstral coefficients and first- and second-order difference coefficients, calculating the cepstral spectrum through linear prediction coefficients, extracting the fundamental vibration frequency of each frame, and using CNN+LSTM to directly input the original speech frame or Mel spectrogram, outputting a high-dimensional feature vector through multi-layer nonlinear transformation to obtain voiceprint features.

[0048] If the user is not logging in for the first time, when the user performs operations on the terminal after logging in, the system will acquire the user's voice and voiceprint features. Based on the user-voiceprint feature table, the system will determine whether there is a risk of identity theft. If so, an alert will be issued and the operation page will be closed. If not, the system will determine whether the user's operation is abnormal. If the operation is abnormal, an alert will be issued and the operation page will be closed. If the operation is normal, the system will continue to detect the user's various operations and determine whether the user's operation is deviated from the expected path. If the operation is deviated from the expected path, the system will determine whether the user's operation is abnormal. If the operation is abnormal, an alert will be issued and the operation page will be closed.

[0049] The process of determining whether a user's operation is abnormal is as follows: When a user performs an operation, the user's voice, operation actions, and operation time of each operation action are acquired in real time. Based on the user's real-time voice, each voice command is acquired, and the operation actions that the user needs to complete in each group and the operation order of each operation action in each group are predicted. Each group of operation actions is called a control group, and the operation actions in each control group are sorted according to the operation order of each operation action.

[0050] The system retrieves all completed user actions and categorizes them into a detection group. It then sorts the actions within the detection group and compares them with those in control groups. If an action in the detection group matches an action in a control group, the system retrieves the action time and the user's age group to determine if any actions exceed the time limit. If so, the user's action is considered abnormal; otherwise, it's considered normal. If all actions in the detection group differ from those in any control group, the user's action is also considered abnormal.

[0051] It should be noted that, based on the user's age group, the operation time required for each operation of the user is obtained from the voice-operation relationship diagram. The actual operation time of each operation of the user is compared with the operation time required for each operation. If the actual operation time of an operation is less than the required operation time, it means that the operation has not timed out. If the actual operation time of an operation is longer than the required operation time, it means that the operation has timed out.

[0052] The specific process for predicting the user's required actions and the order of actions within each group is as follows: The user's voice commands are acquired and divided into voice detection groups. These voice commands are then sorted chronologically. The number of voice commands in each group is determined, and sub-voice groups are identified based on this number. Each voice command in the detection group is compared sequentially with the voice commands in its sub-voice groups. Voice groups whose voice commands match those in the detection group are designated as marked voice groups. The operation groups within each marked voice group are then identified from the voice-operation relationship graph and used as marked operation groups. These marked operation groups represent the user's required actions, and the order of actions within each marked operation group is the order of actions within each group.

[0053] It should be noted that, assuming there are three voice commands in a voice detection group, the first three voice commands of each voice group are obtained, and the first three voice commands of each voice group are the sub-voice groups of each voice group. This example is for illustrative purposes only and is not the only limitation.

[0054] The specific process for determining whether a user's operation deviates from the above is as follows: obtain each operation and each completed operation from each control group of the user, and remove the completed operation from each control group to obtain each sub-control group.

[0055] The system will continue to monitor the user's actions, which will be referred to as marked actions. These actions will be arranged in chronological order, and each marked action will be compared with the actions in each sub-control group. If each marked action is the same as the actions in a sub-control group, it means that the user's actions have not deviated. If each marked action is different from the actions in each sub-control group, it means that the user's actions have deviated.

[0056] The warning terminal is used to issue a warning and close the operation page when the user is not the user or when the user's operation is abnormal.

[0057] The database is used to store the operation actions, the operation sequence of each operation action, the operation duration of each voice command and operation action, and the age of each historical user in each operation process retrieved from the database.

[0058] This invention employs a microphone array installed at a suitable location to collect user voice data and identify various voice commands. Simultaneously, it retrieves real-time operation data and voice recordings from a database, along with the ages of historical users, to construct a voice-operation relationship graph. When a user performs an operation, the system detects the risk of identity theft. If such a risk exists, a warning is issued and the operation page is closed. If not, the system determines whether the user's operation is abnormal. If abnormal, a warning is issued and the operation page is closed, ensuring the accuracy of abnormal operation detection and the effectiveness of operation timeout judgment.

[0059] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. A video IVR risk identification system based on a call center, characterized in that, Includes the following modules: The data acquisition module is used to record the user's operation data in real time when the user performs an operation, and to select a suitable location to install a microphone array at that location to collect the user's voice in real time through the microphone array; The data analysis module is used to retrieve the operation actions and voice commands of each user in each historical operation process from the database, and to construct a voice-operation relationship diagram based on the operation actions and voice commands of each user in each historical operation process. The risk identification module is used to obtain the personal information entered by the user and the user's real-time voice when the user performs an operation, and to determine whether there is a risk of identity theft. If there is, an alert is issued and the operation page is closed. If not, the module determines whether the user's operation is abnormal based on the user's real-time voice, real-time operation data and voice-operation relationship diagram. If the user's operation is abnormal, an alert is issued and the operation page is closed. The warning terminal is used to issue a warning and close the operation page when the user is not the user or when the user's operation is abnormal; The database is used to store the actions, the order of actions, the duration of voice commands and actions, and the age of each user in the past operations.

2. The video IVR risk identification system based on a call center according to claim 1, characterized in that, The specific process for selecting a suitable location is as follows: Obtain the operation terminal page display, and set each mounting point on the border of the operation terminal page display according to the preset distance interval, and install a microphone array at each mounting point. When the user performs voice operation in the operation area, the microphone array at each mounting point collects the user's voice. The signal strength and signal-to-noise ratio of the user's voice collected by the microphone array at each installation point are obtained and compared with the target signal strength and target signal-to-noise ratio. The installation points of each microphone array with a signal strength greater than the target signal strength and a signal-to-noise ratio greater than the target signal-to-noise ratio are identified and referred to as each marked installation point. Obtain the angle between each marker installation point and the user, and compare it with a preset angle threshold. Marker installation points whose angle with the user is less than the preset angle threshold are called backup installation points. Randomly select one backup installation point from the backup installation points. This backup installation point is a suitable installation position for the microphone array.

3. The video IVR risk identification system based on a call center according to claim 1, characterized in that, The data analysis module operates as follows: Each historical user is referred to as a marked user. The operation actions, the operation sequence of each operation action, and each voice command in each historical operation process of each marked user are obtained from the database. The relationship between each voice command and each operation action is analyzed. The ages of each tagged user are retrieved from the database and compared. Users whose ages are in the same age range are grouped into a user group. This method is used to obtain each user group. Within each user group, and based on each age group, each marked user is divided into a user group. Within each user group, the operation actions and operation durations of each operation action in the historical operation process of each marked user are obtained from the database. The operation durations of each operation action of users in different age groups are analyzed. Based on the relationship between each voice command and the sequence of each operation action, and the operation time of each operation action for users of different age groups, a voice-operation relationship diagram is constructed.

4. The video IVR risk identification system based on a call center according to claim 3, characterized in that, The specific process of analyzing the relationship between each voice command and each operation action is as follows: Each historical operation process of each marked user is referred to as a marked operation process. Based on each voice command in each marked operation process, the similarity value between each marked operation process is obtained. Each marked operation process with a similarity value of 1 is divided into an operation process group. Each operation process group is obtained in this way. In each operation process group, obtain each operation action and the operation sequence of each operation action in each marked operation process, and compare them. Marked operation processes with the same operation actions and the same operation sequence are divided into a sub-operation process group. In this way, each sub-operation process group of each operation process group is obtained. Each voice command corresponding to each operation process group is divided into a voice group, and each operation action corresponding to each sub-operation process group of each operation process group and the order of each operation action are divided into each operation group. Then, there is a voice operation relationship between each voice command in the voice group of each operation process group and each operation action in each operation group of each operation group.

5. A video IVR risk identification system based on a call center according to claim 4, characterized in that, The specific process for obtaining the similarity value between each marking operation is as follows: The voice commands of each marking operation process are compared. If the voice commands of multiple marking operation processes are the same and the order of the voice commands is also the same, then the similarity value between the multiple marking operation processes is 1. If the voice commands of multiple marking operation processes are different or the order of the voice commands is different, then the similarity value between the multiple marking operation processes is 0.

6. A video IVR risk identification system based on a call center according to claim 3, characterized in that, The analysis of the operation time of various actions by users of different age groups is as follows: In each user group, each operation action in each marked operation process is obtained. Identical operation actions are grouped into a marked group. Each marked group is obtained in this way, and each marked operation process in each marked group is obtained. The operation time of the operation action corresponding to each marked group in each marked operation process is obtained and compared to obtain the maximum operation time of each operation action. The maximum operation time of each operation action in each user group is obtained in this way, and it is used as the operation time of each operation action of each user group corresponding to the age group. The operation time of each operation action of users in different age groups is obtained in this way.

7. A video IVR risk identification system based on a call center according to claim 1, characterized in that, The risk identification module operates as follows: When a user logs in, it is determined whether the user is logging in for the first time. If it is the first time, a face detection is performed during the login process to determine whether there is a risk of identity theft. If there is a risk, an alert is issued and the operation page is closed. If there is no risk of identity theft, the user is prompted to read a text set by the system, and the user's voice is collected to obtain the user's voiceprint characteristics. The user's personal information and voiceprint characteristics are recorded in the user-voiceprint characteristic table in the database. If the user is not logging in for the first time, when the user performs operations on the terminal after logging in, the system will acquire the user's voice and voiceprint features. Based on the user-voiceprint feature table, the system will determine whether there is a risk of identity theft. If so, an alert will be issued and the operation page will be closed. If not, the system will determine whether the user's operation is abnormal. If the operation is abnormal, an alert will be issued and the operation page will be closed. If the operation is normal, the system will continue to detect the user's various operations and determine whether the user's operation is deviated from the expected path. If the operation is deviated from the expected path, the system will determine whether the user's operation is abnormal. If the operation is abnormal, an alert will be issued and the operation page will be closed.

8. A video IVR risk identification system based on a call center according to claim 7, characterized in that, The specific process for determining whether a user's operation is abnormal is as follows: When a user performs an operation, the system acquires the user's voice, operation actions, and operation time of each operation action in real time. Based on the user's real-time voice, it acquires each voice command, predicts each set of operation actions that the user needs to complete, and the operation order of each operation action in each set of operation actions. Each set of operation actions is called a control group, and the operation actions in each control group are sorted according to the operation order of each operation action. The system retrieves all completed user actions and categorizes them into a detection group. It then sorts the actions within the detection group and compares them with those in control groups. If an action in the detection group matches an action in a control group, the system retrieves the action time and the user's age group to determine if any actions exceed the time limit. If so, the user's action is considered abnormal; otherwise, it's considered normal. If all actions in the detection group differ from those in any control group, the user's action is also considered abnormal.

9. A video IVR risk identification system based on a call center according to claim 8, characterized in that, The specific process of predicting the sets of operations that the user needs to complete and the order of operations within each set of operations is as follows: The system acquires the user's voice commands and divides them into voice detection groups. The voice commands within each detection group are then sorted chronologically. The number of voice commands in each detection group is calculated, and based on this number, sub-voice groups are identified for each voice group. Each voice command in the detection group is then compared sequentially with each voice command in one of its sub-voice groups. Voice groups whose sub-voice commands match those in the detection group are designated as marked voice groups. The system then extracts the operation groups for each marked voice group from the voice-operation relationship graph and designates them as marked operation groups. These marked operation groups represent the user's required actions, and the order of operations within each marked operation group corresponds to the order of operations within each group.

10. A video IVR risk identification system based on a call center according to claim 8, characterized in that, The specific process for determining whether the user's operation has deviated is as follows: Obtain each operation action and each completed operation action from each control group of the user, and remove the completed operation actions from each control group to obtain each sub-control group; The user's actions will continue to be monitored and are referred to as marked actions. These actions will be arranged in chronological order, and each marked action will be compared with the actions in each sub-control group. If each marked action is the same as the actions in a sub-control group, it means that the user's actions have not deviated. If each marked action is different from the actions in each sub-control group, it means that the user's actions have deviated.

Images