Social relation graph-based anthropomorphic vehicle-mounted voice assistant system and interaction method

Abstract

The invention relates to an anthropomorphic vehicle-mounted voice assistant system based on a social relation graph and an interaction method. By endowing the voice assistant with emotional and anthropomorphic characteristics, a social relation graph between the voice assistant and the vehicle owner and the passenger is established. The vehicle-mounted voice assistant system comprises a voice assistant role setting module, a passenger relation recognition and association module, a personalized preference learning and memorizing module, a diversified interaction style module and a social relation graph construction module. According to the invention, the voice assistant is endowed with emotional and anthropomorphic characteristics, so that the defects of the existing vehicle-mounted voice assistant in a multi-person scene are improved; emotional interaction is realized by establishing a social relation graph between the voice assistant and the vehicle owner and the passenger; the service is personalized; and diversified interaction styles are improved. According to the improvement, the voice assistant is not only a tool, but also an intelligent companion with emotion and personalized characteristics, so that the interactive experience and emotion connection between the user and the voice assistant are greatly enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of in-vehicle voice assistants, and specifically relates to an anthropomorphic in-vehicle voice assistant system based on a social relationship graph and an interaction method thereof. Background Art

[0002] At present, the in-vehicle voice assistants on the market generally have voice recognition and command execution functions, but these functions are often implemented based on fixed voice commands or simple voice interaction modes. Voice assistants cannot conduct personalized interactions based on the social relationships between different users, nor can they customize the language style and interaction mode based on factors such as the user's identity, gender, and age.

[0003] At the same time, existing voice assistants cannot automatically identify the relationships between different passengers in multi-passenger scenarios and adjust the interaction mode based on these relationships. In addition, voice assistants lack the ability to continuously learn user preferences, making it difficult to achieve truly personalized services, especially in scenarios involving complex social relationships such as family members. Existing technologies have the following major pain points:

[0004] 1. Lack of emotion: Voice assistants are usually designed as neutral and mechanical characters, unable to adjust their interaction methods according to different social situations;

[0005] 2. Insufficient personalization: Existing voice assistants lack the ability to remember and learn passengers’ preferences and are unable to provide continuous, personalized advice and care;

[0006] 3. Single interactive experience: Whether facing the elderly or children, the interactive style of the voice assistant is relatively single and lacks specificity.

[0007] Based on this, it is necessary to develop an anthropomorphic in-vehicle voice assistant system based on a social relationship graph to effectively solve the above problems. Summary of the invention

[0008] The purpose of the present invention is to provide an anthropomorphic in-vehicle voice assistant system based on a social relationship graph, and also to provide an interaction method for the above-mentioned anthropomorphic in-vehicle voice assistant system. The present invention gives the voice assistant emotional and anthropomorphic characteristics, establishes a social relationship graph between the voice assistant and the car owner and passengers, so as to solve and improve the problems of the existing in-vehicle voice assistant lacking emotion, insufficient personalization and single interactive experience in multi-person scenarios.

[0009] The objective of the present invention is achieved through the following technical solutions:

[0010] An anthropomorphic in-vehicle voice assistant system based on a social relationship map, comprising a voice assistant role setting module, a passenger relationship recognition and association module, a personalized preference learning and memory module, a diversified interaction style module and a social relationship map construction module;

[0011] The voice assistant role setting module is used to determine the voice assistant's way of addressing the owner, the voice assistant's language style and the interaction method when the vehicle is configured with the voice assistant for the first time or the owner sets the social role;

[0012] The passenger relationship identification and association module is used to automatically identify and store the relationship between the passenger and the vehicle owner when a new passenger gets on the vehicle, based on the voice command input by the vehicle owner to introduce the relationship between the passenger and himself, and to infer the relationship between the voice assistant and the passenger based on the preset roles;

[0013] The personalized preference learning and memory module is used to learn and remember each passenger's preferences through continuous interaction, such as music type, adjustment control, common navigation destinations, etc. The voice assistant will provide personalized recommendations and services based on these preferences in subsequent interactions.

[0014] The diversified interaction style module is used for the voice assistant to adjust its language style and interaction mode according to the age, identity, and relationship between the car owner and the voice assistant of different passengers;

[0015] The social relationship map construction module is used for the voice assistant to construct a social relationship map of the car owner and his passengers through the car owner introduction and passenger identification method, recording the relationship between the car owner and each passenger, as well as the association relationship between the voice assistant and each passenger.

[0016] Furthermore, it also includes a user interface interaction module, a speech synthesis model module and a semantic analysis module. The user interface interaction module is used for the car owner to select and set the role of the voice assistant through the car computer interface; the speech synthesis model can synthesize the corresponding speech model according to the role selection of the car owner; the semantic analysis module can perform semantic analysis on the car owner's input voice commands.

[0017] Furthermore, the roles include system preset roles, or new role configurations created through custom options.

[0018] Furthermore, the voice assistant role setting module includes a role database submodule, in which multiple role options are pre-set, each option is associated with a specific language style, voice intonation, way of address and interactive behavior pattern; the voice assistant role setting module also includes a voice model adjustment submodule, which can automatically adjust the speech synthesis model and semantic analysis according to the owner's choice to match the language characteristics of the selected role.

[0019] Furthermore, the passenger relationship identification and association module includes an NLP submodule that can parse the voice commands of the car owner; it also includes a relationship mapping algorithm submodule and a relationship database submodule, the relationship mapping algorithm submodule is used to automatically infer the relationship between the passenger and the voice assistant; the relationship database submodule is used to store all identified relationships, and the database records each passenger's identity information, relationship with the car owner, and relationship with the voice assistant.

[0020] Furthermore, the personalized preference learning and memory module includes a user behavior monitoring submodule for monitoring and recording passengers' voice commands and behavior data; it also includes a machine learning algorithm submodule and a user profile database submodule. The machine learning algorithm submodule is used to analyze and recognize patterns of passengers' behavior data to generate personalized preference profiles; the user profile database submodule is capable of storing each passenger's preference information in the user profile database for future personalized service calls.

[0021] Furthermore, the information recorded in the user profile database includes preference type, frequency, and time period.

[0022] Furthermore, the diversified interaction style module includes a passenger attribute recognition submodule, which can identify the age group and identity attributes of the passenger through the introduction of the car owner or the direct inquiry of the passenger by the voice assistant; it also includes a language style adjustment module submodule and an interaction scene processing submodule. The language style adjustment module submodule has built-in multiple language styles covering passengers of different age groups and identities; the interaction scene processing submodule can call appropriate language style templates in specific interactions.

[0023] Furthermore, the social relationship graph construction module includes a graph construction algorithm sub-module, which can use a graph structure data model to construct and maintain a social relationship graph of the car owner and its passengers; it also includes a relationship update and management module sub-module and a graph query interface sub-module, the relationship update and management module sub-module is used to support dynamic update and management of the graph; the graph query interface sub-module is used for the voice assistant to query the social relationship graph before each interaction, determine the identity and relationship of the current interactive object, and thus call the appropriate language style and interaction method.

[0024] An interaction method of a personified in-vehicle voice assistant system based on a social relationship graph comprises the following steps:

[0025] A. The car owner selects a social role preset in the role database through the car interface, or creates a new role configuration through the custom option;

[0026] B. Query the role database and automatically adjust the voice assistant's speech synthesis model and semantic analysis based on the car owner's role selection to find the language features that match the selected role;

[0027] C. The NLP submodule parses the voice commands of the car owner, automatically infers the relationship between the passenger and the voice assistant by executing the relationship mapping algorithm, and stores the identified relationship in the relational database;

[0028] D. The voice assistant monitors and records the passengers’ voice commands and behavior data, and uses machine learning algorithms to analyze and recognize patterns in the passengers’ behavior data, thereby generating personalized preference profiles. Each passenger’s preference information will be stored in the user profile database for future personalized service calls;

[0029] E. Through the introduction of the car owner or the direct inquiry of the voice assistant to the passenger, the passenger's age group, identity and other attributes are identified. In the specific interaction, the voice assistant calls the appropriate language style template;

[0030] F. Use graph structure data model to build and maintain social relationship graph between car owners and their passengers;

[0031] G. Update the social relationship map. Before each interaction, the voice assistant queries the social relationship map to determine the identity and relationship of the current interaction object, so as to call the appropriate language style and interaction method.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The present invention improves the shortcomings of existing in-vehicle voice assistants in multi-person scenarios by giving the voice assistant emotional and anthropomorphic characteristics; and achieves the following improvements by establishing a social relationship map between the voice assistant and the car owner and passengers:

[0034] 1. Emotional interaction: By defining the voice assistant as a family member or friend of the car owner, the voice assistant can interact with different passengers in different ways according to the relationship between them, thereby improving the user experience. For example, when the car owner is a woman and defines the voice assistant as "sister", the voice assistant can establish a more intimate interaction by calling the car owner's son "nephew";

[0035] 2. Personalized service: Voice assistants can remember and learn the preferences of each passenger and make personalized recommendations or provide care in future interactions, which not only improves the user experience but also enhances the intelligence level of voice assistants.

[0036] 3. Diverse interaction styles: Depending on the age and identity of the passengers, the voice assistant can use different language styles to interact with them; for example, use respectful language for elders and more lively and interesting language for children;

[0037] The above improvements make the voice assistant not just a tool, but also an "intelligent companion" with emotional and personalized characteristics, greatly enhancing the interactive experience and emotional connection between users and voice assistants. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 Flowchart of the steps of the interaction method of a humanized in-vehicle voice assistant system based on a social relationship graph. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with embodiments:

[0041] The present invention discloses an anthropomorphic in-vehicle voice assistant system based on a social relationship map, which includes a voice assistant role setting module, a passenger relationship recognition and association module, a personalized preference learning and memory module, a diversified interaction style module and a social relationship map construction module, and can realize an emotional and anthropomorphic voice assistant and its diversified interaction functions.

[0042] The voice assistant role setting module is used to allow the owner to choose to assign a specific social role to the voice assistant when the vehicle is first configured with the voice assistant or when setting the role. For example, the owner can set the voice assistant as "sister", "brother", "friend", etc. This role setting not only affects the way the voice assistant addresses the owner, but also determines the voice assistant's language style and interaction method.

[0043] Role database: Multiple role options are pre-set in the system, each of which is associated with a specific language style, voice intonation, way of addressing, and interactive behavior pattern.

[0044] User interface interaction: Car owners can select and set the role of the voice assistant through the car interface. Car owners can choose the system preset role or create a new role configuration through the custom option.

[0045] Voice model adjustment: Based on the car owner's choice, the voice assistant's speech synthesis model and semantic parsing module will automatically adjust to match the language characteristics of the selected character. For example, after selecting the "sister" character, the voice assistant's voice may be younger and more lively, and use an intimate way of addressing.

[0046] Through the role setting module, the voice assistant of the present invention can be closer to the life scenes of the car owner and enhance the intimacy and emotional connection of the interaction.

[0047] The passenger relationship identification and association module is used to allow the owner to introduce the relationship between the passenger and himself through voice commands when a new passenger gets on the bus. The voice assistant automatically identifies and stores the relationship between the passenger and the owner through these inputs, and infers the relationship between the voice assistant and the passenger based on the preset roles.

[0048] Natural Language Processing (NLP) submodule: The NLP submodule built into the voice assistant can interpret the owner's voice commands. For example, when the owner says "This is my son", the NLP submodule will recognize the relational word "son" and associate it with the owner.

[0049] Relationship mapping algorithm: The system has a built-in relationship mapping algorithm to automatically infer the relationship between the passenger and the voice assistant. Assuming that the voice assistant is set as the owner's "sister", then through relationship mapping, the system will automatically identify the passenger as the voice assistant's "nephew".

[0050] Relationship database: All identified relationships will be stored in a relationship database to facilitate subsequent personalized interactions. The database records each passenger’s identity information, relationship with the car owner, and relationship with the voice assistant.

[0051] This module enables the voice assistant to automatically adjust the interaction method based on the relationship between the car owner and the passenger, thereby enhancing the personalization and emotionality of the interaction.

[0052] The personalized preference learning and memory module is used for the voice assistant to learn and remember each passenger's preferences through continuous interaction, such as music type, adjustment control, common navigation destinations, etc. The voice assistant will provide personalized recommendations and services based on these preferences in subsequent interactions.

[0053] User behavior monitoring: Voice assistants monitor and record passengers’ voice commands and behavior data, such as frequently played songs, frequently visited places, etc. By analyzing this data, the system can identify passengers’ preferences.

[0054] Machine learning algorithm: The system uses machine learning algorithms to analyze and identify patterns in passenger behavior data, thereby generating personalized preference profiles. For example, if a passenger plays a certain type of music multiple times, the system will mark it as a "preferred music type" and give priority to recommending music of that type in subsequent interactions.

[0055] User profile database: Each passenger's preference information will be stored in the user profile database for future personalized service calls. The database records information including preference type, frequency, time period, etc.

[0056] By learning and remembering passengers' preferences, voice assistants can provide more considerate and personalized services, improving user experience and satisfaction.

[0057] The diversified interaction style module is used for the voice assistant to adjust its language style and interaction mode according to the age, identity, and relationship between the car owner and the voice assistant of different passengers. For example, more respectful language is used for elders, and more lively and interesting language is used for children.

[0058] Passenger attribute identification: Through the introduction of the car owner or the direct inquiry of the voice assistant to the passenger, the system can identify the passenger's age group, identity and other attributes. For example, after identifying that the passenger is an elderly person, the system will mark him / her as "elder".

[0059] Language style adjustment submodule: The system has built-in multiple language style templates, covering passengers of different ages and identities. For example, the elders template will use more respectful titles and tones, while the children template will contain more lively and encouraging language.

[0060] Interaction scenario processing: In specific interactions, the voice assistant will call the appropriate language style template. For example, when interacting with elders, the voice assistant will use "you" and avoid using overly casual words; when interacting with children, the voice assistant may adopt a more exaggerated tone and use words such as "fun".

[0061] Diverse interaction styles enable voice assistants to better adapt to the needs and preferences of different passengers, improving the naturalness of interaction and user satisfaction.

[0062] The social relationship map construction module is used for the voice assistant to construct a social relationship map of the car owner and his passengers through car owner introduction, passenger identification, etc. This map not only records the relationship between the car owner and each passenger, but also records the association relationship between the voice assistant and each passenger.

[0063] Graph construction algorithm: The system uses a graph structure data model to build and maintain a social relationship graph between the car owner and his passengers. Each node represents a passenger, and the edge represents the relationship between them. The attributes of the node include identity, age, relationship with the car owner, etc.

[0064] Relationship update and management: The graph supports dynamic update and management. For example, when the car owner reintroduces the same passenger on different occasions, the system will update the information of the relevant relationship nodes. The graph also supports the addition, deletion and modification of passengers.

[0065] Graph query interface: Before each interaction, the voice assistant will query the social relationship graph to determine the identity and relationship of the current interaction object, so as to call the appropriate language style and interaction method.

[0066] The social relationship graph enables voice assistants to systematically manage and utilize passengers' relationship information, thereby providing more intelligent and emotional interactive services in complex multi-person riding scenarios.

[0067] The present invention can significantly improve the emotionality and anthropomorphism of the voice assistant and provide a more intelligent and personalized user experience.

[0068] In the present invention, the voice assistant's emotional and anthropomorphic role setting: giving the voice assistant a social role related to the car owner, such as a family member or friend, thereby enhancing the emotional color of the interaction. Automatic recognition and memory of passenger relationships: through the car owner's introduction, the voice assistant can automatically recognize and remember the relationship between the passenger, the car owner and the voice assistant. Personalized preference learning and memory: the voice assistant can learn and remember the passenger's preferences and provide continuous personalized recommendations. Diversified interactive language styles: adjust the voice assistant's language style and interaction method according to the passenger's age and identity.

[0069] In the present invention, the construction and application of social relationship graph: the technical solution of protecting the voice assistant to achieve emotional interaction by constructing the social relationship graph between the car owner and the passengers and themselves. Personalized learning algorithm: the algorithm of protecting the voice assistant to perform personalized learning and recommendation based on the passenger behavior data. Diversified interaction style: the technical solution of protecting the voice assistant to adjust the language style according to the passenger's age, identity, and social role.

[0070] like Figure 1 As shown, the interaction method of the anthropomorphic in-vehicle voice assistant system based on the social relationship graph of the present invention includes the following steps:

[0071] A. The car owner selects a social role preset in the role database through the car interface, or creates a new role configuration through the custom option;

[0072] B. Query the role database and automatically adjust the voice assistant's speech synthesis model and semantic analysis based on the car owner's role selection to find the language features that match the selected role;

[0073] C. The NLP submodule parses the voice commands of the car owner, automatically infers the relationship between the passenger and the voice assistant by executing the relationship mapping algorithm, and stores the identified relationship in the relational database;

[0074] D. The voice assistant monitors and records the passengers’ voice commands and behavior data, and uses machine learning algorithms to analyze and recognize patterns in the passengers’ behavior data, thereby generating personalized preference profiles. Each passenger’s preference information will be stored in the user profile database for future personalized service calls;

[0075] E. Through the introduction of the car owner or the direct inquiry of the voice assistant to the passenger, the passenger's age group, identity and other attributes are identified. In the specific interaction, the voice assistant calls the appropriate language style template;

[0076] F. Use graph structure data model to build and maintain social relationship graph between car owners and their passengers;

[0077] G. Update the social relationship map. Before each interaction, the voice assistant queries the social relationship map to determine the identity and relationship of the current interaction object, so as to call the appropriate language style and interaction method.

[0078] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A humanized in-vehicle voice assistant system based on a social relationship graph, characterized in that: It includes voice assistant role setting module, passenger relationship recognition and association module, personalized preference learning and memory module, diversified interaction style module and social relationship map construction module; The voice assistant role setting module is used to determine the voice assistant's way of addressing the owner, the voice assistant's language style and the interaction method when the vehicle is configured with the voice assistant for the first time or the owner sets the social role; The passenger relationship identification and association module is used to automatically identify and store the relationship between the passenger and the vehicle owner when a new passenger gets on the vehicle, based on the voice command input by the vehicle owner to introduce the relationship between the passenger and himself, and to infer the relationship between the voice assistant and the passenger based on the preset roles; The personalized preference learning and memory module is used to learn and remember each passenger's preferences through continuous interaction, such as music type, adjustment control, common navigation destinations, etc. The voice assistant will provide personalized recommendations and services based on these preferences in subsequent interactions. The diversified interaction style module is used for the voice assistant to adjust its language style and interaction mode according to the age, identity, and relationship between the car owner and the voice assistant of different passengers; The social relationship map construction module is used for the voice assistant to construct a social relationship map of the car owner and his passengers through the car owner introduction and passenger identification method, recording the relationship between the car owner and each passenger, as well as the association relationship between the voice assistant and each passenger.

2. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 1, characterized in that: It also includes a user interface interaction module, a speech synthesis model module and a semantic analysis module. The user interface interaction module is used for the car owner to select and set the role of the voice assistant through the car computer interface; the speech synthesis model can synthesize the corresponding speech model according to the car owner's role selection; the semantic analysis module can perform semantic analysis on the car owner's input voice commands.

3. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 2, characterized in that: The roles include system preset roles, or new role configurations created through custom options.

4. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 1, characterized in that: The voice assistant role setting module includes a role database submodule, in which multiple role options are pre-set, each option is associated with a specific language style, voice intonation, way of address and interactive behavior pattern; the voice assistant role setting module also includes a voice model adjustment submodule, which can automatically adjust the speech synthesis model and semantic analysis according to the owner's choice to match the language characteristics of the selected role.

5. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 1, characterized in that: The passenger relationship identification and association module includes an NLP submodule that can parse the voice commands of the car owner; it also includes a relationship mapping algorithm submodule and a relationship database submodule. The relationship mapping algorithm submodule is used to automatically infer the relationship between the passenger and the voice assistant; the relationship database submodule is used to store all identified relationships. The database records the identity information of each passenger, the relationship with the car owner, and the relationship with the voice assistant.

6. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 1, characterized in that: The personalized preference learning and memory module includes a user behavior monitoring submodule for monitoring and recording the voice instructions and behavior data of passengers; a machine learning algorithm submodule and a user profile database submodule, wherein the machine learning algorithm submodule is used to analyze and perform pattern recognition on the behavior data of passengers, thereby generating a personalized preference profile; The user profile database submodule can store the preference information of each passenger in the user profile database for future personalized service calls.

7. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 6, characterized in that: The information recorded in the user profile database includes preference type, frequency, and time period.

8. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 1, characterized in that: The diversified interaction style module includes a passenger attribute recognition submodule, which can identify the age group and identity attributes of the passenger through the introduction of the car owner or the direct inquiry of the voice assistant to the passenger; it also includes a language style adjustment module submodule and an interaction scene processing submodule. The language style adjustment module submodule has built-in multiple language styles covering passengers of different age groups and identities; the interaction scene processing submodule can call appropriate language style templates in specific interactions.

9. The humanized in-vehicle voice assistant system based on social relationship graph according to claim 1, characterized in that: The social relationship graph construction module includes a graph construction algorithm submodule, which can use a graph structure data model to construct and maintain a social relationship graph of the car owner and its passengers; it also includes a relationship update and management module submodule and a graph query interface submodule. The relationship update and management module submodule is used to support dynamic update and management of the graph; the graph query interface submodule is used for the voice assistant to query the social relationship graph before each interaction to determine the identity and relationship of the current interactive object, so as to call the appropriate language style and interaction method.

10. The interactive method of a humanized in-vehicle voice assistant system based on a social relationship graph according to claim 1, characterized in that: The following steps are involved: A. The car owner selects a social role preset in the role database through the car interface, or creates a new role configuration through the custom option; B. Query the role database and automatically adjust the voice assistant's speech synthesis model and semantic analysis based on the car owner's role selection to find the language features that match the selected role; C. The NLP submodule parses the voice commands of the car owner, automatically infers the relationship between the passenger and the voice assistant by executing the relationship mapping algorithm, and stores the identified relationship in the relational database; D. The voice assistant monitors and records the passengers’ voice commands and behavior data, and uses machine learning algorithms to analyze and recognize patterns in the passengers’ behavior data, thereby generating personalized preference profiles. Each passenger’s preference information will be stored in the user profile database for future personalized service calls; E. Through the introduction of the car owner or the direct inquiry of the voice assistant to the passenger, the passenger's age group, identity and other attributes are identified. In the specific interaction, the voice assistant calls the appropriate language style template; F. Use graph structure data model to build and maintain social relationship graph between car owners and their passengers; G. Update the social relationship map. Before each interaction, the voice assistant queries the social relationship map to determine the identity and relationship of the current interaction object, so as to call the appropriate language style and interaction method.

Images