Method, device, equipment and storage medium for interaction between AI virtual human and user

Abstract

The present invention relates to the field of computer technology, and provides an AI virtual person and user interaction method, device, equipment and storage medium, the method comprising: constructing a user instruction pool and a virtual person behavior pool, establishing a mapping correspondence between the user instruction pool and the virtual person behavior pool; constructing an AI virtual person and multiple virtual scenes, establishing a mapping correspondence between the user instruction pool and the virtual scene; obtaining the target user's real-time instruction; obtaining the target user's surrounding environment data, and correcting the target user's real-time instruction according to the target user's surrounding environment data; determining the virtual scene based on the corrected target user's real-time instruction, and importing the AI ​​virtual person into the determined virtual scene; in the determined virtual scene, controlling the AI ​​virtual person to output the virtual person behavior matching the corrected target user's real-time instruction to the target user. The present invention can improve the anthropomorphic interaction effect and intelligence of the AI ​​virtual person.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, equipment and storage medium for interaction between an AI virtual person and a user. Background Art

[0002] With the development of virtual human intelligence and image materialization technology, the use of virtual human display is becoming more and more extensive. Virtual human refers to a virtual character with a digital appearance, which will rely on display devices to exist and have human appearance, human behavior and human thoughts (talking with people).

[0003] After searching, the applicant found some typical prior arts, such as the Chinese invention patent with application number CN202210894618.1, which discloses a human-computer interaction system based on virtual human technology, which decomposes the virtual human into several model base parts, and improves the response speed of the virtual human and the immediacy of the interaction by reasonably arranging the adjustment order of the model base parts. Another example is the Chinese invention patent with application number CN202211044475.1, which discloses a virtual human behavior interaction system. Compared with the previous process-based virtual human interaction method, it adds more interactive response adjustment effects, and further improves the anthropomorphic interaction effect of the virtual human. Another example is the Chinese invention patent with application number CN202410041457.0, which discloses an AI virtual human interaction method and device, which can provide users with more complex and diversified interactions to meet the user's interaction needs.

[0004] It can be seen that there are still many practical problems that need to be dealt with in the actual application of virtual human interaction methods, and there are still many specific solutions that have not been proposed. Summary of the invention

[0005] One object of the present invention is to provide an AI virtual human and user interaction method, which can correct the real-time instructions of a target user according to the surrounding environment data of the target user and determine a virtual scene based on the corrected real-time instructions of the target user, so that the virtual human outputs virtual human behaviors matching the corrected real-time instructions of the target user toward the target user in the determined virtual scene, thereby improving the anthropomorphic interaction effect of the AI ​​virtual human.

[0006] The method for interaction between an AI virtual person and a user comprises the following steps:

[0007] Constructing a user instruction pool and a virtual human behavior pool, and establishing a mapping correspondence between the user instruction pool and the virtual human behavior pool according to a preset first mapping list;

[0008] Constructing an AI virtual person and a plurality of virtual scenes, and establishing a mapping correspondence between the user instruction pool and the virtual scenes according to a preset second mapping list;

[0009] Obtain real-time instructions from target users;

[0010] Acquire the target user's surrounding environment data, and modify the target user's real-time instruction according to the target user's surrounding environment data;

[0011] Determine a virtual scene based on the corrected real-time instruction of the target user, and import the AI ​​virtual person into the determined virtual scene;

[0012] In the determined virtual scene, controlling the AI ​​virtual human to output a virtual human behavior matching the modified real-time instruction of the target user toward the target user;

[0013] The virtual human behavior pool includes a body movement data set, a voice dialogue data set and a facial expression data set, and the user instruction pool includes a body movement instruction set, a voice instruction set and a facial expression instruction set.

[0014] The method for interacting with an AI virtual human and a user constructs a user instruction pool, a virtual human behavior pool, an AI virtual human, and a plurality of virtual scenes, modifies the target user's real-time instructions according to the target user's surrounding environment data, and determines the virtual scene based on the modified target user's real-time instructions, so that the virtual human outputs virtual human behaviors that match the modified target user's real-time instructions to the target user in the determined virtual scene, taking into account the possible influence of the target user's surrounding environment on the target user's real-time instructions, and the modified target user's real-time instructions are more in line with the target user's real intentions, and can improve the anthropomorphic interaction effect and intelligence level of the AI ​​virtual human.

[0015] Preferably, the specific method for obtaining the target user's real-time instruction includes the following steps:

[0016] Identify potential users in the surrounding environment and obtain the instruction level coefficient of each potential user ;

[0017] The potential user with the highest instruction level coefficient is determined as the target user, and the real-time instruction of the potential user with the highest instruction level coefficient is used as the real-time instruction of the target user;

[0018] in, Indicates the spatial distance between the potential user and the preset virtual human coordinate point, Indicates the degree of instruction pointing of potential users, Indicates the correlation between the instruction content of the potential user and the subject attribute, , are the weight values ​​of the distance from the space and the degree of instruction pointing, respectively, Indicates potential users The standard value of the attribute, Indicates potential users The actual value of the attribute, express and The correlation function between .

[0019] Preferably, the method for interaction between an AI virtual person and a user further comprises the following steps:

[0020] Determine whether the target user stops outputting real-time instructions within a preset time period. If not, continue to control the AI ​​virtual human to output virtual human behaviors that match the corrected real-time instructions of the target user toward the target user. If so, update the potential user with the second highest instruction level coefficient to the target user, and use the real-time instructions of the potential user with the second highest instruction level coefficient as the real-time instructions of the target user.

[0021] Preferably, the and The correlation function between .in, is the number of potential user attributes.

[0022] Preferably, the specific method for correcting the target user's real-time instruction according to the target user's surrounding environment data comprises the following steps:

[0023] Acquire potential user data characteristics and objective environment data characteristics according to the target user's surrounding environment data;

[0024] Generate instruction constraints according to the potential user data characteristics and objective environment data characteristics;

[0025] It is determined whether the target user's real-time instruction meets the instruction constraint condition. If not, the target user's real-time instruction is modified.

[0026] Preferably, the specific method for modifying the target user's real-time instruction includes: replacing the target user's real-time instruction with a standard instruction that has the greatest similarity to the target user's real-time instruction and meets the instruction constraint condition, thereby modifying the target user's real-time instruction.

[0027] Preferably, the specific method for determining the virtual scene based on the modified real-time instruction of the target user comprises the following steps:

[0028] Acquire the modified instruction feature of the target user's real-time instruction;

[0029] Calculate the virtual scene matching index based on the potential user data characteristics, objective environment data characteristics and instruction characteristics , ;

[0030] Determine the virtual scene according to the real-time virtual scene matching index;

[0031] in, Indicates Potential user data features about The weight value of a virtual scene, Indicates Objective environmental data features about The weight value of a virtual scene, Indicates The instruction features are about The weight value of a virtual scene, represents the weight adjustment coefficient corresponding to the potential user data characteristics, Represents the weight adjustment coefficient corresponding to the objective environmental data characteristics, represents the weight adjustment coefficient corresponding to the instruction feature, represents the number of potential user data features, Represents the number of objective environmental data features, Indicates the number of instruction features, Indicates the number of virtual scenes.

[0032] The present invention also provides an AI virtual human and user interaction device, which comprises:

[0033] A first mapping module, used to construct a user instruction pool and a virtual human behavior pool, and establish a mapping correspondence between the user instruction pool and the virtual human behavior pool according to a preset first mapping list;

[0034] A second mapping module is used to construct an AI virtual person and multiple virtual scenes, and establish a mapping correspondence between the user instruction pool and the virtual scenes according to a preset second mapping list;

[0035] An instruction acquisition module is used to acquire real-time instructions from target users;

[0036] A correction module, used for acquiring the surrounding environment data of the target user, and correcting the real-time instruction of the target user according to the surrounding environment data of the target user;

[0037] A virtual scene importing module, used to determine a virtual scene based on the corrected real-time instruction of the target user, and import the AI ​​virtual person into the determined virtual scene;

[0038] A control module, used for controlling the AI ​​virtual human to output a virtual human behavior matching the modified real-time instruction of the target user toward the target user in the determined virtual scene;

[0039] The virtual human behavior pool includes a body movement data set, a voice dialogue data set and a facial expression data set, and the user instruction pool includes a body movement instruction set, a voice instruction set and a facial expression instruction set.

[0040] The present invention also provides an AI virtual human and user interaction device, which includes:

[0041] Controller;

[0042] A memory storing executable instructions;

[0043] Among them, the executable instructions can be run on the controller and implement the AI ​​virtual human and user interaction method.

[0044] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program implements the method for interaction between an AI virtual person and a user when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0046] Figure 1 It is a schematic diagram of the overall process of a method for interaction between an AI virtual person and a user in one embodiment of the present invention;

[0047] Figure 2 This is a partial flow chart of a method for interaction between an AI virtual person and a user in one embodiment of the present invention;

[0048] Figure 3 It is a partial flow chart of a method for interaction between an AI virtual human and a user in another embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with its embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the protection scope of the present invention.

[0050] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0052] The “first” and “second” mentioned in the present invention do not represent specific quantities and orders, but are merely used to distinguish names.

[0053] With the development of virtual human intelligence and image materialization technology, the current scenarios for virtual human display and use are becoming more and more extensive. When a virtual human interacts with a user, it will first collect the user's real-time instructions such as voice instructions, facial expressions or body movements, and then output the corresponding content to the user based on the preset virtual human interaction logic and behavior control. The user's surrounding environment often affects the accuracy of the virtual human's collection of user instructions, which in turn affects the accuracy of the virtual human's interaction with the user.

[0054] In order to solve the above problems, Figure 1 As shown, a method for interaction between an AI virtual person and a user in one embodiment of the present invention includes the following steps:

[0055] S1, constructing a user instruction pool and a virtual human behavior pool, and establishing a mapping correspondence between the user instruction pool and the virtual human behavior pool according to a preset first mapping list.

[0056] Specifically, the user instruction pool includes a body movement instruction set, a voice instruction set and a facial expression instruction set. The virtual human behavior pool includes a body movement data set, a voice dialogue data set and a facial expression data set.

[0057] In the preset first mapping list, one user instruction is mapped to at least one specific behavior of a virtual person.

[0058] S2, constructing an AI virtual person and multiple virtual scenes, and establishing a mapping correspondence between the user instruction pool and the virtual scenes according to a preset second mapping list.

[0059] In the preset second mapping list, one user command mapping corresponds to one virtual scene.

[0060] The preset first mapping list and the preset second mapping list can be preset by a technician and will not be described in detail here.

[0061] S3, obtaining the target user's real-time instructions. The target user's real-time instructions include body movement instructions, voice instructions, and facial expression instructions.

[0062] The target user includes at least one, and the target user is determined by collecting the identity characteristics of the user in the environment and comparing and matching with the pre-stored identity characteristics of the target user. In some cases, the target user can be determined by collecting the face image of the user in the environment, recognizing the face image, and then matching the preset face image of the target user with the recognized face image.

[0063] S4, obtaining the target user's surrounding environment data, and modifying the target user's real-time instruction according to the target user's surrounding environment data.

[0064] In some cases, the target user's surrounding environment data may affect the target user's real-time instructions obtained, such as environmental noise and / or the voices of non-target users will interfere with the collection of the target user's voice instructions, or the light brightness will affect the collection of the target user's facial expression instructions. By obtaining the target user's surrounding environment data and correcting the target user's real-time instructions, real-time instructions that are more in line with the target user's true intentions can be obtained, thereby improving the intelligence and accuracy of the interaction between the virtual human and the user.

[0065] S5, determining a virtual scene based on the corrected real-time instruction of the target user, and importing the AI ​​virtual person into the determined virtual scene.

[0066] A default virtual scene is set among the multiple virtual scenes, and when a corresponding virtual scene cannot be determined based on the corrected real-time instruction of the target user, the default virtual scene is determined as the virtual scene corresponding to the corrected real-time instruction of the target user.

[0067] Importing the AI ​​virtual person into the determined virtual scene so that the AI ​​virtual person outputs corresponding content to the target user in the determined virtual scene can enhance the interaction effect.

[0068] S6, in the determined virtual scene, controlling the AI ​​virtual human to output virtual human behaviors matching the corrected real-time instructions of the target user toward the target user.

[0069] The method for interacting with an AI virtual human and a user constructs a user instruction pool, a virtual human behavior pool, an AI virtual human, and a plurality of virtual scenes, modifies the target user's real-time instructions according to the target user's surrounding environment data, and determines the virtual scene based on the modified target user's real-time instructions, so that the virtual human outputs virtual human behaviors that match the modified target user's real-time instructions to the target user in the determined virtual scene, taking into account the possible influence of the target user's surrounding environment on the target user's real-time instructions, and the modified target user's real-time instructions are more in line with the target user's real intentions, and can improve the anthropomorphic interaction effect and intelligence level of the AI ​​virtual human.

[0070] In one embodiment, if Figure 2 As shown, in step S3, the specific method for obtaining the target user's real-time instruction includes the following steps:

[0071] S30, determining potential users in the surrounding environment and obtaining the instruction level coefficient of each potential user .

[0072] Potential users refer to users who may issue instructions to interact with virtual people. Preferably, potential users in the surrounding environment can be determined by identifying the real-time location and / or identity characteristics of the users.

[0073] S31, determining the potential user with the highest instruction level coefficient as the target user, and taking the real-time instruction of the potential user with the highest instruction level coefficient as the real-time instruction of the target user.

[0074] in, Indicates the spatial distance between the potential user and the preset virtual human coordinate point, Indicates the degree of instruction pointing of potential users, Indicates the correlation between the instruction content of the potential user and the subject attribute, , are the weight values ​​of the distance from the space and the degree of instruction pointing, which can be set according to the actual situation. Indicates potential users The standard value of the attribute, Indicates potential users The actual value of the attribute, express and The correlation function between .

[0075] The preset virtual human coordinate points may be the spatial position coordinate points of electronic devices (such as servers and 3D projectors) used to display virtual humans and virtual scenes. Specifically, the spatial coordinates of real scenes such as halls, guest rooms, or exhibition rooms may be constructed, and then the specific spatial coordinates of the electronic devices may be used as the preset virtual human coordinate points.

[0076] The instruction pointing degree refers to the possibility that the instruction issued by the potential user is directed to the virtual person. Specifically, the key information of the potential user's instruction can be extracted, and the instruction pointing degree of the potential user can be obtained based on the key information. For example, if the potential user's instruction is a voice instruction, the instruction pointing degree can be obtained by extracting the keywords in the voice instruction. More specifically, different keywords can be preset to correspond to different instruction pointing degrees. Similarly, different body movement features and facial expression features can also preset corresponding instruction pointing degrees. That is to say, in the potential user's instruction, different key information corresponds to different preset instruction pointing degrees.

[0077] Preferably, the and The correlation function between .in, is the number of potential user attributes. Potential user attributes include but are not limited to height, age, expression, and voice. and The correlation function between is used to identify the identity of the user. By pre-storing different attribute standard values ​​of multiple potential users, extracting the actual attribute values ​​of the potential users in real time, calculating the correlation between the attribute standard value of the potential user and the actual attribute value, and then determining the identity of the potential user. The correlation is between -1 and 1. The larger the absolute value, the stronger the linear relationship between the attribute standard value of the potential user and the actual attribute value, and the more consistent the potential user in the surrounding environment is with the pre-stored potential user's identity.

[0078] The subject attributes of the potential user are identified by voiceprint feature recognition, image feature recognition, body shape feature recognition, or a combination of at least two of the voiceprint feature recognition, image feature recognition, and body shape feature recognition, and the subject identity of the potential user is determined. The subject identity includes but is not limited to the elderly, children, women, students, adults, and infants.

[0079] The correlation between the instruction content of the potential user and the subject attribute is obtained by extracting the key information of the instruction content of the potential user and calculating the correlation between the key information and the subject attribute. Preferably, for different subject attributes, multiple key information are preset and corresponding correlation values ​​are preset for different key information. After determining the subject identity of the potential user, the correlation between the instruction content of the potential user and the subject attribute is obtained by calculating the average value or weighted average value of the key information.

[0080] By obtaining the instruction level coefficient of each potential user , taking into account the spatial distance between the potential user and the preset virtual human coordinate point, the correlation between the instruction content and the subject attributes, and the degree of instruction pointing, and then determining the potential user with the highest instruction level coefficient as the target user, and taking the real-time instructions of the potential user with the highest instruction level coefficient as the real-time instructions of the target user, the priority of the potential user's instructions is divided, which can realize the interaction between the virtual human and the user more intelligently and further improve the interaction effect.

[0081] In one embodiment, the method for interaction between the AI ​​virtual person and the user further includes the following steps:

[0082] Determine whether the target user stops outputting real-time instructions within a preset time period. If not, continue to control the AI ​​virtual human to output virtual human behaviors that match the corrected real-time instructions of the target user toward the target user. If so, update the potential user with the second highest instruction level coefficient to the target user, and use the real-time instructions of the potential user with the second highest instruction level coefficient as the real-time instructions of the target user.

[0083] In some cases, such as when there are multiple potential users in a real space scene, multiple potential users may issue different instructions at the same time. Their instruction level coefficients are different, but they are all within the scope of the virtual person's ability to interact with the user based on the instruction content. When the instruction level coefficient is the highest, that is, when the target user has completed outputting real-time instructions and no longer continues to output real-time instructions, the potential user with the second highest instruction level coefficient is updated as the target user, and the real-time instructions of the potential user with the second highest instruction level coefficient are used as the target user's real-time instructions, which can enhance the interaction effect between the virtual person and the user, so that the virtual person can continue to interact with the user based on the instruction content of the potential user.

[0084] Preferably, after the virtual person uses the real-time instruction of the potential user with the second highest instruction level coefficient as the target user real-time instruction and completes outputting the contents of a preset number of real-time instructions, steps S30 and S31 may be re-executed.

[0085] In many cases, different users need different content to interact with virtual humans. Some interactive content is suitable for one user, but may not be suitable for other users. In addition, the content output by virtual humans in different objective environments also needs to be subject to certain constraints. Even for the same user, the interactive content that the virtual human outputs indoors is sometimes not suitable for display outdoors or in other places. For this reason, it is necessary to modify the real-time instructions of the target user according to the differences in potential users and actual places in the virtual human and user interaction environment to ensure that the interactive content output by the virtual human will not have a negative impact on others.

[0086] In order to solve the above problem, in one embodiment, as Figure 3 As shown, in step S4, the specific method for correcting the target user's real-time instruction according to the target user's surrounding environment data includes the following steps:

[0087] S40, acquiring data characteristics of potential users and objective environment data characteristics according to the surrounding environment data of the target user.

[0088] S41, generating instruction constraints according to the potential user data characteristics and the objective environment data characteristics.

[0089] S42, judging whether the target user's real-time instruction meets the instruction constraint condition, and if not, modifying the target user's real-time instruction.

[0090] Preferably, in step S42, the specific method for correcting the target user's real-time instruction includes: replacing the target user's real-time instruction with a standard instruction that has the greatest similarity to the target user's real-time instruction and meets the instruction constraint condition, thereby correcting the target user's real-time instruction.

[0091] The standard instructions are pre-configured in the virtual person. The potential user data features include but are not limited to potential user group features, such as the subject identity of the potential user (including the elderly, children, women, students, adults and infants). Objective environment data features include but are not limited to objective environment sound features and image features.

[0092] The group categories included in the potential users in the objective environment are determined according to the data characteristics of the potential users, and the categories of the objective environment (such as indoor or outdoor, community, school or shopping mall, etc.) are determined according to the data characteristics of the objective environment. Then, instruction constraints are generated according to the group categories included in the potential users and the objective environment categories. The method for generating instruction constraints here can specifically be: sensitive information is obtained according to the group categories included in the potential users and the objective environment categories, and instruction constraints are generated according to the sensitive information. Among them, sensitive information corresponding to different group categories and different objective environment categories is pre-configured in the virtual person.

[0093] When the key information in the target user's real-time instruction matches the sensitive information, it is determined that the target user's real-time instruction does not meet the instruction constraint condition.

[0094] By modifying the real-time instructions of the target user, the content output by the virtual person is made to meet the instruction constraints, thereby avoiding the virtual person from outputting sensitive information and causing discomfort to others.

[0095] In one embodiment, in step S5, the specific method of determining the virtual scene based on the modified real-time instruction of the target user comprises the following steps:

[0096] S50, obtaining the modified instruction feature of the target user's real-time instruction.

[0097] S51, calculating the virtual scene matching index according to the potential user data characteristics, objective environment data characteristics and instruction characteristics ， 。

[0098] S52: Determine a virtual scene according to the real-time virtual scene matching index.

[0099] in, Indicates Potential user data features about The weight value of a virtual scene, Indicates Objective environmental data features about The weight value of a virtual scene, Indicates The instruction features are about The weight value of a virtual scene, represents the weight adjustment coefficient corresponding to the potential user data characteristics, Represents the weight adjustment coefficient corresponding to the objective environmental data characteristics, represents the weight adjustment coefficient corresponding to the instruction feature, represents the number of potential user data features, Represents the number of objective environmental data features, Indicates the number of instruction features, Indicates the number of virtual scenes.

[0100] Here, by pre-configuring multiple different virtual scenes in the electronic device and configuring corresponding weight values ​​for different potential user data features, instruction features and objective environment data features, the virtual scene matching indexes of the multiple different virtual scenes are calculated, and the virtual scene with the highest virtual scene matching index is used as the virtual scene that matches the modified real-time instruction of the target user.

[0101] Combining the data characteristics of potential users, objective environment characteristics and instruction characteristics to determine the virtual scene, and importing the virtual person into the determined virtual scene, the virtual person can meet the potential user group, objective environment and real-time instruction content of the target user to the greatest extent, thereby enhancing the interaction effect.

[0102] In one of the embodiments, the present invention also provides an AI virtual human and user interaction device, which includes a first mapping module, a second mapping module, an instruction acquisition module, a correction module, a virtual scene import module and a control module.

[0103] The first mapping module is used to construct a user instruction pool and a virtual human behavior pool, and establish a mapping correspondence between the user instruction pool and the virtual human behavior pool according to a preset first mapping list; the second mapping module is used to construct an AI virtual human and multiple virtual scenes, and establish a mapping correspondence between the user instruction pool and the virtual scenes according to a preset second mapping list.

[0104] The instruction acquisition module is used to acquire the target user's real-time instructions; the correction module is used to acquire the target user's surrounding environment data and correct the target user's real-time instructions according to the target user's surrounding environment data.

[0105] The virtual scene importing module is used to determine the virtual scene based on the corrected real-time instructions of the target user, and import the AI ​​virtual human into the determined virtual scene; the control module is used to control the AI ​​virtual human to output virtual human behaviors matching the corrected real-time instructions of the target user toward the target user in the determined virtual scene.

[0106] The virtual human behavior pool includes a body movement data set, a voice dialogue data set and a facial expression data set, and the user instruction pool includes a body movement instruction set, a voice instruction set and a facial expression instruction set.

[0107] The AI ​​virtual human and user interaction device constructs a user instruction pool, a virtual human behavior pool, an AI virtual human and multiple virtual scenes, modifies the target user's real-time instructions according to the target user's surrounding environment data, and determines the virtual scene based on the modified target user's real-time instructions, so that the virtual human outputs virtual human behaviors that match the modified target user's real-time instructions to the target user in the determined virtual scene, taking into account the possible impact of the target user's surrounding environment on the target user's real-time instructions. The modified target user's real-time instructions are more in line with the target user's real intentions, and can improve the anthropomorphic interaction effect and intelligence level of the AI ​​virtual human.

[0108] In one of the embodiments, the present invention also provides an AI virtual human and user interaction device, which includes: a controller; and a memory storing executable instructions.

[0109] Among them, the executable instructions can be run on the controller and implement the AI ​​virtual human and user interaction method.

[0110] In one of the embodiments, the present invention further provides a computer-readable storage medium storing a computer program, wherein the computer program implements the method for interaction between an AI virtual human and a user when executed by a processor.

[0111] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. The method for interaction between AI virtual human and user is characterized by: The method for interaction between an AI virtual person and a user comprises the following steps: Constructing a user instruction pool and a virtual human behavior pool, and establishing a mapping correspondence between the user instruction pool and the virtual human behavior pool according to a preset first mapping list; Constructing an AI virtual person and a plurality of virtual scenes, and establishing a mapping correspondence between the user instruction pool and the virtual scenes according to a preset second mapping list; Obtain real-time instructions from target users; Acquire the target user's surrounding environment data, and modify the target user's real-time instruction according to the target user's surrounding environment data; Determine a virtual scene based on the corrected real-time instruction of the target user, and import the AI ​​virtual person into the determined virtual scene; In the determined virtual scene, controlling the AI ​​virtual human to output a virtual human behavior matching the modified real-time instruction of the target user toward the target user; The virtual human behavior pool includes a body movement data set, a voice dialogue data set, and a facial expression data set, and the user instruction pool includes a body movement instruction set, a voice instruction set, and a facial expression instruction set; The specific method of obtaining the target user's real-time instructions includes the following steps: Identify potential users in the surrounding environment and obtain the instruction level coefficient of each potential user ; The potential user with the highest instruction level coefficient is determined as the target user, and the real-time instruction of the potential user with the highest instruction level coefficient is used as the real-time instruction of the target user; in, Indicates the spatial distance between the potential user and the preset virtual human coordinate point, Indicates the degree of instruction pointing of potential users, Indicates the correlation between the instruction content of the potential user and the subject attribute, , are the weight values ​​of the distance from the space and the degree of instruction pointing, respectively, Indicates potential users The standard value of the attribute, Indicates potential users The actual value of the attribute, express and The correlation function between The specific method for correcting the target user's real-time instruction according to the target user's surrounding environment data comprises the following steps: Acquire potential user data characteristics and objective environment data characteristics according to the target user's surrounding environment data; Generate instruction constraints according to the potential user data characteristics and objective environment data characteristics; Determine whether the target user's real-time instruction meets the instruction constraint condition, and if not, modify the target user's real-time instruction; The group category included in the potential user in the objective environment is determined according to the potential user data characteristics, the category of the objective environment is determined according to the objective environment data characteristics, and then the instruction constraint condition is generated according to the group category included in the potential user and the objective environment category; the method for generating the instruction constraint condition is specifically: sensitive information is obtained according to the group category included in the potential user and the objective environment category, and the instruction constraint condition is generated according to the sensitive information.

2. The AI ​​virtual human and user interaction method according to claim 1, characterized in that: The method for interaction between AI virtual human and user also includes the following steps: Determine whether the target user stops outputting real-time instructions within a preset time period. If not, continue to control the AI ​​virtual human to output virtual human behaviors that match the corrected real-time instructions of the target user toward the target user. If so, update the potential user with the second highest instruction level coefficient to the target user, and use the real-time instructions of the potential user with the second highest instruction level coefficient as the real-time instructions of the target user.

3. The AI ​​virtual human and user interaction method according to claim 2, characterized in that: Said and The correlation function between ,in, is the number of potential user attributes.

4. The AI ​​virtual human and user interaction method according to claim 3, characterized in that: The specific method for modifying the target user's real-time instruction includes: replacing the target user's real-time instruction with a standard instruction that has the greatest similarity to the target user's real-time instruction and meets the instruction constraint condition, thereby modifying the target user's real-time instruction.

5. The AI ​​virtual human and user interaction method according to claim 4, characterized in that: The specific method for determining the virtual scene based on the modified target user real-time instruction comprises the following steps: Acquire the modified instruction feature of the target user's real-time instruction; Calculate the virtual scene matching index based on the potential user data characteristics, objective environment data characteristics and instruction characteristics , ; Determine the virtual scene according to the real-time virtual scene matching index; in, Indicates Potential user data features about The weight value of a virtual scene, Indicates Objective environmental data features about The weight value of a virtual scene, Indicates The instruction features are about The weight value of a virtual scene, represents the weight adjustment coefficient corresponding to the potential user data characteristics, Represents the weight adjustment coefficient corresponding to the objective environmental data characteristics, represents the weight adjustment coefficient corresponding to the instruction feature, represents the number of potential user data features, Represents the number of objective environmental data features, Indicates the number of instruction features, Indicates the number of virtual scenes.

6. The AI ​​virtual human and user interaction device is characterized by: The AI ​​virtual person and user interaction device comprises: A first mapping module, used to construct a user instruction pool and a virtual human behavior pool, and establish a mapping correspondence between the user instruction pool and the virtual human behavior pool according to a preset first mapping list; A second mapping module is used to construct an AI virtual person and multiple virtual scenes, and establish a mapping correspondence between the user instruction pool and the virtual scenes according to a preset second mapping list; An instruction acquisition module is used to acquire real-time instructions from target users; A correction module, used for acquiring the surrounding environment data of the target user, and correcting the real-time instruction of the target user according to the surrounding environment data of the target user; A virtual scene importing module, used to determine a virtual scene based on the corrected real-time instruction of the target user, and import the AI ​​virtual person into the determined virtual scene; A control module, used for controlling the AI ​​virtual human to output a virtual human behavior matching the modified real-time instruction of the target user toward the target user in the determined virtual scene; The virtual human behavior pool includes a body movement data set, a voice dialogue data set, and a facial expression data set, and the user instruction pool includes a body movement instruction set, a voice instruction set, and a facial expression instruction set; The specific method of obtaining the target user's real-time instructions includes the following steps: Identify potential users in the surrounding environment and obtain the instruction level coefficient of each potential user ; The potential user with the highest instruction level coefficient is determined as the target user, and the real-time instruction of the potential user with the highest instruction level coefficient is used as the real-time instruction of the target user; in, Indicates the spatial distance between the potential user and the preset virtual human coordinate point, Indicates the degree of instruction pointing of potential users, Indicates the correlation between the instruction content of the potential user and the subject attribute, , are the weight values ​​of the distance from the space and the degree of instruction pointing, respectively, Indicates potential users The standard value of the attribute, Indicates potential users The actual value of the attribute, express and The correlation function between The specific method for correcting the target user's real-time instruction according to the target user's surrounding environment data comprises the following steps: Acquire potential user data characteristics and objective environment data characteristics according to the target user's surrounding environment data; Generate instruction constraints according to the potential user data characteristics and objective environment data characteristics; Determine whether the target user's real-time instruction meets the instruction constraint condition, and if not, modify the target user's real-time instruction; The specific method for correcting the target user's real-time instruction according to the target user's surrounding environment data comprises the following steps: Acquire potential user data characteristics and objective environment data characteristics according to the target user's surrounding environment data; Generate instruction constraints according to the potential user data characteristics and objective environment data characteristics; Determine whether the target user's real-time instruction meets the instruction constraint condition, and if not, modify the target user's real-time instruction; The group category included in the potential user in the objective environment is determined according to the potential user data characteristics, the category of the objective environment is determined according to the objective environment data characteristics, and then the instruction constraint condition is generated according to the group category included in the potential user and the objective environment category; the method for generating the instruction constraint condition is specifically: sensitive information is obtained according to the group category included in the potential user and the objective environment category, and the instruction constraint condition is generated according to the sensitive information.

7. AI virtual human and user interaction device, characterized in that: The AI ​​virtual person and user interaction device includes: Controller; A memory storing executable instructions; The executable instructions can be run on the controller and implement the AI ​​virtual human and user interaction method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for interaction between an AI virtual human and a user as described in any one of claims 1 to 5 is implemented.

Images