Digital human interaction management system based on artificial intelligence

Abstract

The present invention discloses a digital human interaction management system based on artificial intelligence, which relates to the field of human-computer interaction technology, and comprises an interaction interface module, a digital human interaction module, an interaction information pre-storage module, a real-time data analysis module, an intelligent analysis module and an interaction information storage module; the interaction interface module is used to obtain user identity information, environmental information during digital human interaction and Internet information, and sets the data authority value that can be called by different users to C, and confirms the current user identity information by connecting with the intelligent analysis module, and obtains the size of the data authority threshold C that can be called by different users; the digital human interaction module is used to obtain dialogue information in the digital human interaction process, convert the obtained dialogue information into digital information, and transmit the digital information to the interaction information pre-storage module; avoid bypassing the shielding of privacy words by replacing keywords, so as to obtain the privacy data of different users.

Description

Technical Field

[0001] The present invention relates to the field of human-computer interaction technology, and in particular to a digital human interaction management system based on artificial intelligence. Background Art

[0002] With the continuous breakthroughs in science and technology and the accelerated evolution of artificial intelligence, the digital human industry is welcoming unprecedented development opportunities. From film and television entertainment to marketing, from e-commerce live broadcasting to financial services, digital humans have penetrated into various industries in multiple forms, reshaping the business ecology and user experience; in the context of the rapid development of AI technology, the team has established competitive advantages in multiple key areas with its deep technical accumulation and innovation capabilities. With the development of deep learning and augmented reality technology, virtual assistants have begun to develop in a more intelligent and immersive direction. Augmented reality technology allows users to interact with virtual assistants in a more intuitive and immersive way through devices such as AR glasses;

[0003] Existing artificial intelligence often has multiple users, and there is a risk of privacy leakage during the human-computer interaction process. Existing technologies protect privacy by disabling keywords to prevent different users from calling other people's saved private data during the human-computer interaction process. Existing privacy protection is too procedural and it is easy to obtain different users' private data by replacing keywords.

[0004] In view of the lack of corresponding privacy management in the existing artificial intelligence digital human interaction management process, a solution is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a digital human interaction management system based on artificial intelligence to solve the technical defects mentioned above. The present invention processes voice data, associates and stores voice data, allows the database to be quickly called, reduces the time for data association retrieval, and supports cross-calling of data, improves human-computer interactivity and intelligence, and conducts privacy assessment of voice data through intelligent analysis. By identifying the user's identity, the scope and authority of data called during the human-computer interaction process are limited, thereby realizing the management of privacy data of different users during the interaction process.

[0006] The object of the present invention can be achieved by the following technical solutions: A digital human interaction management system based on artificial intelligence includes an interaction interface module, a digital human interaction module, an interaction information pre-storage module, a real-time data analysis module, an intelligent analysis module and an interaction information storage module;

[0007] The interactive information pre-storage module is used to store and pre-process the acquired digital information. After receiving the voice signal, it is divided into time windows of equal intervals, and the short-time energy is calculated for each time window. The short-time energy is the total energy of the voice signal in each time window. The pre-processed digital signal is connected to the real-time data analysis module and the intelligent analysis module, and the voice signal part from the starting point to the end point is extracted for subsequent voice recognition, instruction parsing and processing. The short-time energy refers to the total energy of the voice signal in each time window.

[0008] The real-time data analysis module is used to analyze the text data of the digital human interaction process in real time. By acquiring the relevance of natural language text data and combining it with the environmental information and Internet information during the digital human interaction, the module stores the data in the database through persistent information transmission, quickly processes the voice data, establishes preliminary data association, and transmits the analyzed and associated data to the interactive information storage module. The start and end points of the voice text are segmented, and the segmented voice text is annotated, and the object name, time and place of the text are annotated.

[0009] The intelligent analysis module is used to analyze the acquired digital information and evaluate the privacy risk of the voice data generated by the user during the digital human interaction process through information entropy. The information entropy is the average amount of information carried by each symbol in the information source. The information entropy is compared with the size of the user's data permission value C to divide the open range of digital information of different users, and the data is transmitted to the interactive information storage module. The entropy is compared by evaluating the privacy risk of the corresponding attributes.

[0010] Preferably, the process of converting the acquired dialogue information into digital information by the digital human interaction module is as follows:

[0011] It is used to obtain the dialogue information during the digital human interaction process, convert the obtained dialogue information into digital information, and transmit the digital information to the interaction information pre-storage module; through multiple audio receiving components at different spatial positions, the audio information at different spatial positions in the digital human interaction process is obtained respectively, and through the adjustment of phase and amplitude, the audio receiving components at different spatial positions can form a beam in a specific direction, and the minimum mean square error criterion is used to optimize the weight of the spatial filter to suppress interference signals.

[0012] Preferably, the energy threshold acquisition process of the interactive information pre-storage module is as follows:

[0013] By converting audio information into text data, estimating the change trend of the signal, and adjusting the threshold according to the estimated value, the basic iterative formula is as follows:

[0014]

[0015] in, is the parameter vector, is the parameter vector at the previous time point, is the parameter vector at the previous time point, For time, is the error covariance matrix, is the error covariance matrix at the previous time point, is the observation vector, is the actual observed value, T is the adjusted energy threshold, determines the starting and ending points of the voice command, extracts the voice signal part from the starting point to the ending point, and calculates it by summing the square of the signal amplitude.

[0016] Preferably, the process of determining the start and end points of the speech signal in the interactive information pre-storage module is as follows:

[0017] The speech signal is , at a length of In the window of The short-term energy of a time window The calculation formula is as follows:

[0018]

[0019] in, is the time index, is the window index, is the window length, the energy threshold , used to determine whether the speech signal starts or ends:

[0020] When short-term energy Exceeding the threshold When , the starting point of the speech signal is considered to be at the starting position of the current window;

[0021] When short-term energy Below or equal to threshold When , and the energy of n consecutive windows is lower than or equal to the threshold , then the end point of the speech signal is determined to be the end position of the current window.

[0022] Preferably, the analysis process of the real-time data analysis module is as follows:

[0023] The acquired digital information is converted into speech text, and the starting and ending points of the speech text are segmented. At the same time, the segmented speech text is annotated, the object name, time and place of the text are annotated, and the data is transmitted to the interactive information storage module for storage. The RNN unit is represented as:

[0024]

[0025] in, It's time The hidden state of It's time The input vector is and is the weight matrix, is the activation function, which describes the hidden state through the nonlinear function ReLU Based on the hidden state of the previous time step and the input of the current time step It is calculated that RNN can retain the first ten lines of information and the last ten lines of information when processing sequence data, making the hidden state Based on the hidden state of the previous time step and the input of the current time step The correlation of the voice data is calculated.

[0026] Preferably, the analysis process of the intelligent analysis module is as follows:

[0027] The obtained interactive information is analyzed and the information entropy is used to evaluate the privacy risk of user attributes. The information entropy of user attributes is calculated by the following formula:

[0028]

[0029] in, represents information entropy, Represents the name attribute, Indicates that the name attribute takes a certain value probability.

[0030] Preferably, the privacy risk assessment process of the intelligent analysis module is as follows:

[0031] The entropy is compared by evaluating the privacy risk of the corresponding attributes. When there are two attributes A and B, when the attribute value A is greater than B, the privacy risk of A is greater than B, and the privacy risk of A is given more importance; otherwise, the privacy risk of B is given more importance.

[0032] By setting the permission threshold C; when the call entropy value range is greater than the permission threshold C, the call of the object user's related text data is rejected;

[0033] When the call entropy value range is less than or equal to C, the call object user-related text data is allowed.

[0034] Preferably, the storage process in the interaction information storage module is as follows:

[0035] It is used to obtain data after real-time data analysis and data after intelligent analysis, and store them. By obtaining the type and structure of interactive information data that needs to be stored and analyzing data access patterns and requirements, the scalability requirements of the system are determined. According to the system requirements and data characteristics, a suitable NoSQL database type is selected, and the database data model is designed according to the requirements. Storage and indexing of data are considered to support the query requirements of the system. At the same time, it is configured according to the requirements of the production environment, database security measures are set, existing data is imported into the NoSQL database, and necessary data structures and initial data are initialized.

[0036] Preferably, the operation process of the interactive interface module is as follows:

[0037] It is used to obtain user identity information, environmental information during digital human interaction, and Internet information, set the data permission value that different users can call to C, and confirm the current user identity information through connection with the intelligent analysis module, and obtain the size of the data permission value C that different users can call.

[0038] Preferably, the authority threshold determination process of the interactive interface module is as follows:

[0039] By obtaining the identity recognition data from the outside world, matching it with the pre-stored data of the interaction information currently called, when the identity recognition data from the outside world is the same as the pre-stored data of the interaction information currently called, setting the authority threshold C of the current user to the maximum value;

[0040] When the identity recognition data in the outside world is different from the pre-stored data of the currently called interaction information, the authority threshold C of the current user is set to an intermediate value of 0.5.

[0041] The beneficial effects of the present invention are as follows:

[0042] The present invention makes the interaction process more intelligent by cross-calling the data between different users, and makes preliminary association between the voice data of different users through real-time analysis of voice data, thereby improving the interactivity of digital human, and saves the associated voice data, thereby reducing the time of data calling and improving the reaction speed of voice data in the interaction process of digital human;

[0043] The present invention calls the voice texts of different users. When the calling process involves the privacy data of other users, the privacy assessment is performed on the relevant privacy data through intelligent analysis. By identifying the user identity, the data scope and authority called during the human-computer interaction process are limited, thereby achieving the management of the privacy data of different users during the interaction process, avoiding bypassing the shielding of privacy words by replacing keywords, and thus obtaining the privacy data of different users. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present invention will be further described below in conjunction with the accompanying drawings;

[0045] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Example:

[0048] See also Figure 1 As shown, the present invention is a digital human interaction management system based on artificial intelligence, including an interaction interface module, a digital human interaction module, an interaction information pre-storage module, a real-time data analysis module, an intelligent analysis module and an interaction information storage module;

[0049] The interactive interface module is used to obtain user identity information, digital human interaction environment information and Internet information, set the data permission value that different users can call to C, and confirm the current user identity information through connection with the intelligent analysis module, and obtain the size of the data permission value C that different users can call;

[0050] The operation process of the interactive interface module is as follows: by obtaining the identity recognition data in the outside world and matching it with the pre-stored data of the interactive information currently called, when the identity recognition data in the outside world is the same as the pre-stored data of the interactive information currently called, the authority threshold C of the current user is set to the maximum value;

[0051] When the identity recognition data in the outside world is different from the pre-stored data of the currently called interactive information, the authority threshold C of the current user is set to the middle value 0.5;

[0052] The digital human interaction module is used to obtain the dialogue information during the digital human interaction process, convert the obtained dialogue information into digital information, and transmit the digital information to the interaction information pre-storage module;

[0053] The data interaction conversion process of the digital human interaction module is as follows: through multiple audio receiving components at different spatial positions, the audio information at different spatial positions during the digital human interaction process is obtained respectively. By adjusting the phase and amplitude, the audio receiving components at different spatial positions can form a beam in a specific direction, and the minimum mean square error criterion is used to optimize the weight of the spatial filter to suppress interference signals;

[0054] The interactive information processing process of the interactive information pre-storage module converts audio information into text data, estimates the change trend of the signal, and adjusts the threshold according to the estimated value. The basic iterative formula is as follows:

[0055]

[0056] in, is the parameter vector, is the parameter vector at the previous time point, is the parameter vector at the previous time point, For time, is the error covariance matrix, is the error covariance matrix at the previous time point, is the observation vector, is the actual observed value, T is the adjusted energy threshold, determines the starting and ending points of the voice command, extracts the part of the voice signal from the starting point to the ending point, and calculates it by summing the square of the signal amplitude;

[0057] The speech signal is , at a length of In the window of The short-term energy of a time window The calculation formula is as follows:

[0058]

[0059] in, is the time index, is the window index, is the window length, the energy threshold , used to determine whether the speech signal starts or ends:

[0060] When short-term energy Exceeding the threshold When , the starting point of the speech signal is considered to be at the starting position of the current window; when the short-time energy Below or equal to threshold When , and the energy of n consecutive windows is lower than or equal to the threshold , then determine the end point of the speech signal as the end position of the current window;

[0061] The analysis process of the real-time data analysis module is as follows: the acquired digital information is converted into speech text, and the start and end points of the speech text are segmented. At the same time, the segmented speech text is annotated, the object name, time and place of the text are annotated, and the data is transmitted to the interactive information storage module for storage. The RNN unit is represented as:

[0062]

[0063] in, It's time The hidden state of It's time The input vector is and is the weight matrix, is the activation function, which describes the hidden state through the nonlinear function ReLU Based on the hidden state of the previous time step and the input of the current time step It is calculated that RNN can retain the first ten lines of information and the last ten lines of information when processing sequence data, making the hidden state Based on the hidden state of the previous time step and the input of the current time step Calculate the relevance of voice data;

[0064] The interactive information pre-storage module is used to store and pre-process the acquired digital information. After receiving the digital signal, it is divided into time windows of equal intervals, and the short-time energy is calculated for each time window. The short-time energy is the total energy of the voice signal in each time window. The pre-processed digital signal is connected to the real-time data analysis module and the intelligent analysis module; the voice signal part from the starting point to the end point is extracted for subsequent voice recognition, instruction parsing and processing. The short-time energy refers to the total energy of the voice signal in each time window.

[0065] The real-time data analysis module is used to analyze the text data of the digital human interaction process in real time, by acquiring the relevance of natural language text data, and combining the environmental information and Internet information during the digital human interaction, and storing it in the database through information persistence transmission, quickly processing voice data, establishing preliminary data association, and transmitting the analyzed associated data to the interactive information storage module; the real-time data analysis module is used to convert the acquired digital information into voice text, and segment the start and end points of the voice text, and at the same time mark the segmented voice text, mark the object name, time and place of the text, and transmit the data to the interactive information storage module for storage;

[0066] The intelligent analysis module is used to analyze the acquired digital information and evaluate the privacy risk of the voice data generated by the user during the digital human interaction process through information entropy. Information entropy is the average amount of information carried by each symbol in the information source. The information entropy is compared with the user's data permission value C to divide the open range of different users' digital information and transmit the data to the interactive information storage module; the intelligent analysis module is used to analyze the acquired interactive information and use information entropy to evaluate the privacy risk of user attributes. It is assumed that the information entropy of user attributes can be calculated by the following formula:

[0067]

[0068] in, represents information entropy, Represents the name attribute, Indicates that the name attribute takes a certain value The probability of calling entropy is compared by evaluating the privacy risk of the corresponding attribute. When there are two attributes A and B, when the attribute value A is greater than B, the privacy risk of A is greater than B. At this time, the privacy risk of A is more important. By setting a threshold C; when the call entropy value range is greater than the threshold C, the call of the object user-related text data is rejected; when the call threshold is less than or equal to C, the call of the object user-related text data is allowed; and connected to the interactive interface module, suppose we have a user data set containing the following information:

[0069] User A: (name="Alice", age=25, interest="music")

[0070] User B: (name="Bob", age=30, interests="sports")

[0071] User C: (name="Charlie", age=22, interest="travel")

[0072] Use information entropy to evaluate the privacy risk of name attributes. First, calculate the information entropy of the name attribute: calculate the occurrence probability of each name:

[0073] Alice: 1 / 3

[0074] Bob: 1 / 3

[0075] Charlie: 1 / 3

[0076] Calculating Information Entropy :

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] Information Entropy Indicates the average amount of information of the name attribute. If the information entropy is higher, it means that the name data is more identifiable and unique, thus increasing the user's identifiability in the system;

[0083] An interactive information storage module is used to obtain and store data after real-time data analysis and data after intelligent analysis;

[0084] By setting a threshold C; when the call entropy value range is greater than the threshold C, the call of the object user's related text data is rejected;

[0085] When the call threshold is less than or equal to C, the call of the target user's related text data is allowed;

[0086] The storage process in the interactive information storage module is as follows:

[0087] The interactive information storage module is used to store data after real-time data analysis and intelligent analysis, determine the type and structure of data to be stored, analyze data access patterns and requirements, and determine the scalability requirements of the system. According to the system requirements and data characteristics, select the appropriate NoSQL database type, design the database data model according to the requirements, and consider how to store and index data to support the query requirements of the system, configure according to the requirements of the production environment, set the security measures of the database, import the existing data into the NoSQL database, and initialize the necessary data structure and initial data for the new system.

[0088] The interactive interface module is used to obtain external information, which includes synchronization information during the interaction process, external environment data, Internet data and interface data, and is connected to the intelligent analysis module. It obtains identity recognition data from the outside world and matches it with the pre-stored data of the currently called interactive information.

[0089] The above formulas are obtained by collecting a large amount of data for software simulation and selecting a formula that is close to the real value. The coefficients in the formula are set by technical personnel in this field according to actual conditions. The above are only preferred specific implementations of the present invention, but the protection scope of the present invention is not limited to this. Any technical personnel familiar with the technical field within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, make equivalent replacement or change, which should be covered within the protection scope of the present invention.

Claims

1. A digital human interaction management system based on artificial intelligence, characterized in that: include: Interaction interface module, digital human interaction module, interaction information pre-storage module, real-time data analysis module, intelligent analysis module and interaction information storage module; The digital human interaction module is used to obtain the dialogue information during the digital human interaction process, convert the obtained dialogue information into digital information, and transmit the digital information to the interaction information pre-storage module; through multiple audio receiving components at different spatial positions, respectively obtain the audio information at different spatial positions during the digital human interaction process, adjust the phase and amplitude so that the audio receiving components at different spatial positions form a beam in the direction of audio signal reception, use the minimum mean square error criterion to optimize the weight of the spatial filter to suppress interference signals, and transmit the interaction information to the interaction information pre-storage module; The interactive information pre-storage module is used to store and pre-process the acquired digital information. After receiving the voice signal, it is divided into time windows of equal intervals, the short-time energy of each time window is calculated, and the energy threshold is obtained. The process is as follows: By converting audio information into text data, estimating the change trend of the signal, and adjusting the threshold according to the estimated value, the basic iterative formula is as follows: ； in, is the parameter vector, is the parameter vector at the previous time point, is the parameter vector at the previous time point, For time, is the error covariance matrix, is the error covariance matrix at the previous time point, is the observation vector, is the actual observed value, is the adjusted energy threshold; short-time energy is the total energy of the speech signal in each time window, and the starting and ending points of the speech signal are judged. The process is as follows: By determining the start and end points of the voice command, the voice signal part from the start point to the end point is extracted, and the voice signal is calculated by summing the square of the signal amplitude. , at a length of In the window of The short-term energy of a time window The calculation formula is as follows: ； in, is the time index, is the window index, is the window length, the energy threshold , used to determine whether the speech signal starts or ends: When short-term energy Exceeding the threshold When , the starting point of the speech signal is considered to be at the starting position of the current window; When short-term energy Below or equal to threshold When , and the energy of n consecutive windows is lower than or equal to the threshold , then determine the end point of the voice signal as the end position of the current window; and connect the pre-processed digital signal with the real-time data analysis module and the intelligent analysis module; The real-time data analysis module is used to analyze the text data of the digital human interaction process in real time, convert the acquired digital information into voice text, and segment the start and end points of the voice text. At the same time, the segmented voice text is annotated, the object name, time and place of the text are annotated, and the data is transmitted to the interactive information storage module for storage. The RNN unit is represented as: ； in, It's time The hidden state of It's time The input vector is and is the weight matrix, is the activation function, which describes the hidden state through the nonlinear function ReLU Based on the hidden state of the previous time step and the input of the current time step It is calculated that RNN can retain the first ten lines of information and the last ten lines of information when processing sequence data, making the hidden state Based on the hidden state of the previous time step and the input of the current time step Calculate the relevance of voice data; obtain the relevance of natural language text data, combine it with the environmental information and Internet information during the digital human interaction, store it in the database through information persistence transmission, quickly process voice data, establish preliminary data association, and transmit the analyzed associated data to the interactive information storage module; The intelligent analysis module is used to analyze the acquired digital information and the acquired interactive information, and use information entropy to evaluate the privacy risk of user attributes. The information entropy of user attributes is calculated by the following formula: ； in, represents information entropy, Represents the name attribute, Indicates that the name attribute takes a certain value The privacy risk of the voice data generated by the user in the digital human interaction process is evaluated by information entropy. Information entropy is the average amount of information carried by each symbol in the information source. The information entropy is compared with the size of the user's data permission value C to divide the open range of different users' digital information, and the data is transmitted to the interactive information storage module.

2. The digital human interaction management system based on artificial intelligence according to claim 1 is characterized in that: The privacy risk assessment process of the intelligent analysis module is as follows: The entropy is compared by evaluating the privacy risk of the corresponding attributes. When there are two attributes A and B, when the attribute value A is greater than B, the privacy risk of A is greater than B, and the privacy risk of A is given more importance; otherwise, the privacy risk of B is given more importance. By setting the permission threshold C; when the call entropy value range is greater than the permission threshold C, the call of the object user's related text data is rejected; When the call entropy value range is less than or equal to C, the call object user-related text data is allowed.

3. The digital human interaction management system based on artificial intelligence according to claim 1 is characterized in that: The storage process in the interactive information storage module is as follows: It is used to obtain data after real-time data analysis and data after intelligent analysis, and store them. By obtaining the type and structure of interactive information data that needs to be stored and analyzing data access patterns and requirements, the scalability requirements of the system are determined. According to the system requirements and data characteristics, a suitable NoSQL database type is selected, and the database data model is designed according to the requirements. Storage and indexing of data are considered to support the query requirements of the system. At the same time, it is configured according to the requirements of the production environment, database security measures are set, existing data is imported into the NoSQL database, and the document storage data structure and initial data are initialized.

4. The digital human interaction management system based on artificial intelligence according to claim 1 is characterized in that: The operation process of the interactive interface module is as follows: It is used to obtain user identity information, environmental information during digital human interaction, and Internet information, set the data permission value that different users can call to C, and confirm the current user identity information through connection with the intelligent analysis module, and obtain the size of the data permission value C that different users can call.

5. The digital human interaction management system based on artificial intelligence according to claim 4 is characterized in that: The authority threshold determination process of the interactive interface module is as follows: By obtaining the identity recognition data from the outside world and matching it with the pre-stored data of the interaction information currently called, when the identity recognition data from the outside world is the same as the pre-stored data of the interaction information currently called, the authority threshold C of the current user is set to the maximum value; When the identity recognition data in the outside world is different from the pre-stored data of the currently called interaction information, the authority threshold C of the current user is set to an intermediate value of 0.5.

Images