A method, apparatus and electronic device for creating personalized virtual characters

Abstract

A method, apparatus, and electronic device for creating personalized virtual characters, relating to the field of data processing, are disclosed. The method involves acquiring visual and auditory stimuli reflecting various character elements and displaying these stimuli to the user; acquiring the user's physiological data regarding each character element; analyzing the physiological data to obtain the user's preference features for each character element, and quantifying these preference features into points in a high-dimensional vector space to form a player preference model; generating a candidate character pool containing multiple candidate characters based on a preset character attribute space; mapping the player preference model to the preset character attribute space to obtain the ideal character and its corresponding target parameters; and determining the actual character from the candidate character pool based on the target parameters and the candidate parameters corresponding to each candidate character, and then displaying the actual character to the user. Implementing the technical solution provided in this application improves the user's gaming experience.

Description

Technical Field

[0001] This application relates to the field of data processing, specifically to a method, apparatus, and electronic device for creating personalized virtual characters. Background Technology

[0002] With the rapid development of digital entertainment and interactive media, virtual characters have become a core element in video games, virtual reality, cloud gaming, and other fields. Users increasingly seek characters that can reflect their unique identity and personality in the virtual world. However, existing virtual character creation technologies are mainly based on predefined templates, and players can only choose from a limited number of options when creating a character.

[0003] This reliance on fixed templates limits the degree of personalization for characters. Players cannot fully express their unique personalities and preferences, resulting in virtual characters that may not adequately reflect the player's true intentions and style, significantly impacting the player's gaming experience.

[0004] Therefore, there is an urgent need for a method, device, and electronic device for creating personalized virtual characters. Summary of the Invention

[0005] This application provides a method, apparatus, and electronic device for creating personalized virtual characters, which improves the user's gaming experience.

[0006] A first aspect of this application provides a method for creating a personalized virtual character. The method includes: acquiring visual and auditory stimuli reflecting various character elements, and displaying the visual and auditory stimuli to a user; acquiring physiological data of the user for each character element, including electroencephalogram (EEG) data, electromyography (EMG) data, and electrodermal transfer (EDT) data; analyzing the physiological data to obtain the user's preference features for each character element, and quantifying the preference features into points in a high-dimensional vector space to form a player preference model; generating a candidate character pool containing multiple candidate characters based on a preset character attribute space, wherein each candidate character is determined by a set of candidate parameters; mapping the player preference model to the preset character attribute space to obtain an ideal character and target parameters corresponding to the ideal character; and determining an actual character from the candidate character pool based on the target parameters and the candidate parameters corresponding to each candidate character, and displaying the actual character to the user.

[0007] By employing the aforementioned technical solution, this method acquires physiological feedback data on user interactions with character elements, analyzes user preferences, generates a quantified player preference model, and creates a candidate character pool based on a preset character attribute space. The player preference model is then mapped to the attribute space to obtain the target parameters for the ideal character. Finally, the method matches the actual character from the candidate character pool that best matches the target parameters and recommends it to the user, thus realizing a novel method for creating personalized virtual characters. This method fully utilizes user physiological data to objectively and comprehensively capture users' true preferences for different character elements. Simultaneously, through a series of steps including preference model quantification, candidate character generation, target parameter mapping, and similarity matching, user preferences and character attributes are accurately and efficiently matched. This ensures that the generated actual character best meets the user's personalized needs and expectations, thereby improving the user's gaming experience.

[0008] Optionally, the step of analyzing the physiological data to obtain the user's preference features for each of the role elements specifically includes: constructing a physiological feature vector based on the physiological data; inputting the physiological feature vector into a preset machine learning model, and obtaining a score corresponding to each of the role elements through inference by the preset machine learning model; sorting each of the role elements according to the score to obtain a preset number of target role elements as the preference features.

[0009] By adopting the above technical solution, a feature vector reflecting the user's physiological state is first constructed, and then a preset machine learning model is used to infer the feature vector to obtain the user's preference score for each role element. Based on the score ranking, the most preferred role elements are selected as preference features, and the user's preference information for role elements is accurately and efficiently extracted from physiological data.

[0010] Optionally, the step of constructing a physiological feature vector based on the physiological data specifically includes: preprocessing the EEG data and extracting the EEG rhythm from the EEG data using a bandpass filter; calculating the power spectral density from the EEG rhythm to obtain the power spectral density features of each frequency band; performing full-wave rectification and smoothing on the muscle wave data to extract the root mean square value of muscle contraction; performing baseline correction and standardization on the skin conductance response data to obtain the skin conductance level and skin conductance response amplitude; and constructing a physiological feature vector based on the power spectral density features, the root mean square value, the skin conductance level, and the skin conductance response amplitude.

[0011] By employing the aforementioned technical solutions, this paper details how to extract key features related to user preferences from physiological signals such as EEG, EMG, and EMG, and construct physiological feature vectors, laying a solid data foundation for accurate analysis of user preferences. Specifically, preprocessing and filtering of EEG signals effectively removes interference such as power frequency noise and baseline drift, improving signal quality. Extracting the power spectral density characteristics of EEG rhythms reflects the distribution of EEG energy in different frequency bands, comprehensively characterizing the user's psychological activity levels across multiple dimensions, including attention, emotion, and cognition. Full-wave rectification and smoothing of EMG signals highlights the energy characteristics of muscle contraction, filters out high-frequency noise, and extracts stable root mean square (RMS) EMG values, sensitively reflecting the intensity of the user's emotional response to stimuli. Baseline correction and standardization of EMG signals eliminate individual differences and signal drift, obtaining standardized EMG levels and response amplitudes, accurately characterizing the user's cognitive and emotional arousal states.

[0012] Optionally, quantizing the preference features into points in a high-dimensional vector space to form a player preference model specifically includes: constructing a character element dictionary and numbering each character element in the dictionary according to a preset order to obtain the mapping relationship between the element ID and element name of each character element; extracting the target character elements contained in each preference feature to obtain an element ID list, the element ID list including multiple target element IDs, the target element ID being the element ID corresponding to the target character element; creating a binary vector based on the length of the element ID list, the length of the binary vector being equal to the size of the character element dictionary, the binary vector being an all-zero vector; traversing each target element ID in the element ID list, determining the corresponding position of each target element ID in the binary vector, and setting the value of the corresponding position to 1 to obtain the preference vector corresponding to the preference feature; mapping the preference vector to a preset high-dimensional vector space, and estimating the probability distribution of the mapped preference vector to obtain the player preference model.

[0013] By adopting the above technical solution, a method is proposed to quantify user preference features into points in a high-dimensional vector space to form a player preference model. This achieves a structured and standardized representation of preference information, providing a high-quality preference knowledge source for subsequent preference matching and personalized recommendations. By constructing a character element dictionary, discrete character elements are mapped to unique element IDs, establishing a correspondence between element IDs and element semantics, forming a character element semantic space that facilitates element identification, retrieval, and comparison. Based on this, the character elements contained in the preference features are transformed into a corresponding list of element IDs, and a binary preference vector of equal length is generated. Each dimension represents whether the user prefers that element, forming a concise and compact preference representation.

[0014] Optionally, generating a candidate character pool containing multiple candidate characters based on a preset character attribute space specifically includes: defining the preset character attribute space, which includes an appearance attribute dimension, a personality attribute dimension, and a skill attribute dimension; determining the attribute categories included in each of the appearance attribute dimension, the personality attribute dimension, and the skill attribute dimension; determining candidate attribute parameters corresponding to each attribute category; and combining the candidate attribute parameters in each attribute category to generate candidate parameters corresponding to multiple candidate characters, thereby obtaining the candidate character pool.

[0015] By adopting the above technical solution, a method for generating a diverse pool of candidate characters based on a preset character attribute space is presented. This provides a rich and comprehensive library of alternative materials for personalized character creation, improving creation efficiency and quality. By defining a standardized preset character attribute space covering multiple dimensions such as appearance, personality, and skills, all aspects of the character are depicted, forming a globally unified character description framework that provides a basis and constraints for the generation of candidate characters.

[0016] Optionally, mapping the player preference model to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character specifically includes: mapping the player preference model to each dimension of the preset character attribute space to obtain the user's preference distribution in each attribute dimension; for each attribute dimension, calculating the expected value of the attribute dimension according to the preference distribution, and using the expected value as the target parameter of the ideal character in the corresponding attribute dimension.

[0017] By adopting the above technical solution, a method is proposed to map the player preference model to the preset character attribute space to obtain the target parameters of the ideal character. This achieves a precise and efficient conversion from user preferences to character attributes, and provides a quantitative and operable target guide for personalized character generation.

[0018] Optionally, the step of determining the actual role in the candidate role pool based on the target parameter and the candidate parameters corresponding to each candidate role, and displaying the actual role to the user, specifically includes: for each candidate role in the candidate role pool, extracting the corresponding candidate parameters; calculating the similarity between the candidate parameter vector and the target parameter vector; sorting the candidate roles according to the similarity, obtaining the candidate role with the highest similarity to the target parameter as the actual role, and displaying the actual role to the user.

[0019] By employing the aforementioned technical solution, a method is presented to determine the optimal actual role from a pool of candidate roles and recommend it to the user, based on target parameters and candidate parameters. This method enables the rapid and accurate selection of personalized roles that best match user preferences from a large pool of candidate roles, significantly improving user experience and satisfaction. By extracting the attribute parameter vector of each candidate role and comparing its similarity with the target parameter vector, the matching degree between the candidate role and the ideal role across various attribute dimensions is quantitatively evaluated. Using vector similarity as an evaluation metric comprehensively considers the differences in the values ​​of each attribute parameter, overcoming the limitations of single-attribute matching and achieving comprehensive and fine-grained matching between candidate and target roles. Based on this, all candidate roles are ranked according to their similarity scores, and the role most similar to the ideal role and closest to the user's preferences is selected as the optimal actual role, ensuring the accuracy and personalization of the recommendation results. Simultaneously, the optimal actual role is visually presented to the user, allowing them to intuitively experience the effect of personalized customization, enhancing user engagement and satisfaction.

[0020] A second aspect of this application provides a personalized virtual character creation device, comprising: an acquisition module and a processing module, wherein: the acquisition module is configured to acquire visual and auditory stimulus materials reflecting various character elements, and display the visual and auditory stimulus materials to a user; the acquisition module is further configured to acquire physiological data of the user for each of the character elements, the physiological data including electroencephalogram (EEG) data, electromyogram (EMG) data, and electrodermal response (EDR) data; the processing module is configured to analyze the physiological data to obtain the user's preference features for each of the character elements, and quantify the preference features into points in a high-dimensional vector space to form a player preference model; the processing module is further configured to generate a candidate character pool containing multiple candidate characters based on a preset character attribute space, wherein each candidate character is determined by a set of candidate parameters; the processing module is further configured to map the player preference model onto the preset character attribute space to obtain an ideal character and target parameters corresponding to the ideal character; the processing module is further configured to determine an actual character from the candidate character pool based on the target parameters and the candidate parameters corresponding to each candidate character, and display the actual character to the user.

[0021] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any of the foregoing.

[0022] A fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed, perform the method described in any of the preceding descriptions.

[0023] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] 1. By acquiring users' physiological feedback data on character elements, analyzing user preferences, generating a quantified player preference model, and creating a candidate character pool based on a preset character attribute space, this method maps the player preference model to the attribute space to obtain the target parameters of the ideal character. Then, it matches the actual character closest to the target parameters from the candidate character pool and recommends it to the user, realizing a novel method for creating personalized virtual characters. This method fully utilizes users' physiological data to objectively and comprehensively capture users' true preferences for different character elements. Simultaneously, through a series of steps including preference model quantification, candidate character generation, target parameter mapping, and similarity matching, it accurately and efficiently matches user preferences with character attributes, ensuring that the generated actual character best meets the user's personalized needs and expectations, thus improving the user's gaming experience. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a personalized virtual character creation method disclosed in an embodiment of this application;

[0026] Figure 2 This is a schematic diagram of a module of a personalized virtual character creation device disclosed in an embodiment of this application;

[0027] Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application.

[0028] Explanation of reference numerals in the attached drawings: 201, acquisition module; 202, processing module; 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. Detailed Implementation

[0029] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0030] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0031] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0032] This application provides a method for creating personalized virtual characters, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a personalized virtual character creation method provided in an embodiment of this application. The method is applied to a server, which executes a personalized virtual character creation program. The server can communicate with user devices via wired or wireless networks. The method includes steps S101 to S106, as follows:

[0033] Step S101: Obtain visual and auditory stimuli materials reflecting each character element, and display the visual and auditory stimuli materials to the user.

[0034] In step S101, the server first acquires visual and auditory stimuli materials reflecting various character elements. These materials can come from the server's local material library or be obtained from external material providers. For visual stimuli materials, they may include, but are not limited to, the following categories: 2D or 3D models of the characters, reflecting the characters' basic appearance, body shape, and clothing features; static images such as portraits and character illustrations, showcasing the characters' detailed features and visual style; and animations of the characters' actions and expressions, reflecting the characters' dynamic features such as body language and facial expressions.

[0035] After the server obtains these materials, it processes and handles them appropriately to better present them to users. For visual materials, scaling, cropping, and color adjustments can be performed to adapt to the display interfaces of different devices; for auditory materials, volume balancing and noise reduction can be applied to ensure clear sound quality without background noise.

[0036] Next, the server will display the processed visual and auditory stimuli to the user. The specific display method can be determined based on the user's device type and interaction method. For PC users, the materials can be displayed on a web page or client software, providing interactive methods such as mouse clicks and keyboard operations; for mobile users, the materials can be displayed in an app, providing touch and swipe gesture operations; for VR / AR device users, the materials can be displayed in a virtual environment, providing immersive interaction such as head tracking and gesture recognition.

[0037] Step S102: Obtain the user's physiological data for each role element. The physiological data includes brainwave data, muscle wave data, and skin conductance response data.

[0038] In step S102, the server acquires the user's physiological data for each role element, including electroencephalogram (EEG) data, electromyography (EMG) data, and electrodermal conductance (EDC) data. This physiological data reflects the user's subconscious reactions and emotional changes to different role elements, providing an objective and quantitative basis for subsequent analysis of user preferences.

[0039] To acquire this physiological data, the user first wears specific sensor devices. For electroencephalogram (EEG) data, an EEG helmet or headband is worn, equipped with multiple electrodes to collect weak potential changes on the scalp surface. For electromyography (EMG) data, surface electrodes are attached to specific muscle groups (such as facial muscles, arm muscles, etc.) to detect potential changes during muscle contraction. EMG devices can include the Myo armband, DTS wireless EMG system, etc. For electrodermal response (EDS) data, electrodes are attached to the user's fingers or palms to measure changes in skin conductivity.

[0040] After the user wears the sensor device, the sensor collects the user's physiological data in real time as the user browses different character elements, and transmits it to the server via wired or wireless means. Upon receiving the data, the server performs a series of preprocessing steps. Next, the server synchronizes the physiological data with the materials the user is browsing. Since physiological responses usually have a certain delay, this time window is considered, and physiological data fragments within a certain range before and after the presentation time of the material are extracted as the corresponding physiological responses for that material.

[0041] Step S103: Analyze the physiological data to obtain the user's preference features for each role element, and quantify the preference features into points in a high-dimensional vector space to form a player preference model.

[0042] In step S103, the physiological data is analyzed to obtain the user's preference features for each role element. Specifically, this includes: constructing a physiological feature vector based on the physiological data; inputting the physiological feature vector into a preset machine learning model, and obtaining the score corresponding to each role element through inference by the preset machine learning model; sorting each role element according to the score to obtain a preset number of target role elements as preference features. The player preference model is formed by quantifying preference features into points in a high-dimensional vector space. Specifically, this involves: constructing a character element dictionary and numbering each character element in the dictionary according to a preset order to obtain the mapping relationship between element IDs and element names; extracting target character elements contained in each preference feature to obtain a list of element IDs, which includes multiple target element IDs, each corresponding to a target character element; creating a binary vector based on the length of the element ID list, with the length of the binary vector equal to the size of the character element dictionary, and the binary vector being an all-zero vector; traversing each target element ID in the element ID list, determining the corresponding position of each target element ID in the binary vector, and setting the value of the corresponding position to 1 to obtain the preference vector corresponding to the preference feature; mapping the preference vector to a preset high-dimensional vector space and estimating the probability distribution of the mapped preference vector to obtain the player preference model.

[0043] Specifically, the server constructs a physiological feature vector based on the user's physiological data. This physiological feature vector typically includes various time-domain and frequency-domain indicators, such as the power of different frequency bands in electroencephalography (EEG), the root mean square amplitude of electromyography (EMG), and the average conductivity of electrodermal conductance. These indicators reflect the intensity and pattern of the user's physiological responses to different audiovisual stimuli. Next, the server inputs the physiological feature vector into a pre-defined machine learning model to predict the user's preference for different role elements. The pre-defined machine learning model can be a support vector machine, random forest, or neural network, among others. Through training, these models learn the mapping relationship between physiological features and user preferences, thereby inferring the user's preference score for role elements based on new physiological data.

[0044] For example, in the training samples, if a user's alpha wave energy and electromyography amplitude increase when viewing a certain character element, then that character element is considered a user preference and is marked as a positive sample; conversely, if the user's physiological response is flat and the questionnaire score is low, it is marked as a negative sample. The pre-set machine learning model is trained using such samples, continuously optimizing the model parameters, and ultimately able to predict users' preference scores for character elements with relatively high accuracy.

[0045] Based on the scores predicted by the machine learning model, the server sorts all role elements and selects the top N (preset number) elements with the highest scores as the user's preference features. These preference features reflect the role elements that the user is most interested in at the current state. To further quantify the preference features, the server first constructs a role element dictionary, numbering all possible role elements in a certain order and establishing a mapping relationship between element IDs and element names. Then, the server extracts the target role elements contained in the user's preference features, obtaining a list of corresponding element IDs.

[0046] Based on the list of element IDs, the server creates a zero-based vector of length equal to the dictionary size, called the preference vector. For each target element ID in the list, the server iterates through the preference vector, finds the corresponding position, and sets the value of that position to 1. This results in a high-dimensional sparse vector, where 1 indicates the user prefers the element, and 0 indicates the user has no strong preference. Finally, the server maps the preference vector to a predefined high-dimensional vector space and estimates the probability distribution of the mapped vector to obtain the player preference model. Mapping methods can include Latent Semantic Analysis (LSA), Principal Component Analysis (PCA), etc., which can transform high-dimensional sparse vectors into low-dimensional dense vectors while preserving the semantic relationships between vectors. Probability distribution estimation models the distribution of the preference vector in the vector space; methods can include Gaussian Mixture Models (GMM), Variational Autoencoders (VAE), etc.

[0047] In one possible implementation, constructing a physiological feature vector based on physiological data specifically includes: preprocessing EEG data and extracting EEG rhythms from the EEG data using a bandpass filter; calculating the power spectral density from the EEG rhythms to obtain the power spectral density features of each frequency band; performing full-wave rectification and smoothing on muscle electroencephalogram data to extract the root mean square value of muscle contraction; performing baseline correction and standardization on skin conductance response data to obtain skin conductance level and skin conductance response amplitude; and constructing a physiological feature vector based on the power spectral density features, root mean square value, skin conductance level, and skin conductance response amplitude.

[0048] Specifically, the server needs to perform a series of preprocessing and feature extraction steps on the user's EEG data, EEG data, and skin conductance data to ultimately construct a physiological feature vector. For EEG data, preprocessing is first performed, including removing power line noise, baseline drift, and other interference to improve signal quality. Preprocessing methods can include notch filtering and wavelet transform. Then, the server uses a bandpass filter to extract specific frequency bands from the EEG data to obtain EEG rhythms. EEG rhythms are typically divided into frequency bands such as δ (0.5-4Hz), θ (4-8Hz), α (8-13Hz), β (13-30Hz), and γ (>30Hz). Energy changes in different frequency bands reflect different brain states and cognitive processes. Next, the server performs power spectral density estimation on the signal segments of each EEG rhythm to obtain the power spectral density characteristics (PSD) of each frequency band. PSD describes the energy distribution of the signal at different frequencies, and estimation methods include periodogram methods and Welch methods. PSD (Power Detection Scheme) can quantify the energy characteristics of brainwave rhythms. For example, the PSD of alpha waves reflects the user's level of relaxation, while the PSD of beta waves reflects the user's level of attention.

[0049] For electromyography (EMG) data, the server first performs full-wave rectification on the raw signal, flipping negative potentials to positive potentials, so that the amplitude of the EMG signal is positively correlated with the strength of muscle contraction. Then, the rectified signal is smoothed to remove high-frequency noise, obtaining the envelope of muscle contraction. Smoothing methods include moving average and low-pass filtering. Finally, the server calculates the root mean square (RMS) value of the smoothed signal as an indicator of muscle contraction intensity. The larger the RMS value, the stronger the muscle contraction and the stronger the user's emotional response.

[0050] For skin conductance response (SCL) data, the server first performs baseline correction to remove slow drift in SCL levels, ensuring a stable signal baseline. Correction methods can include high-pass filtering and trend fitting. Then, the corrected signal is standardized, scaling its amplitude to a fixed range for easier comparison between different users. Standardization methods include maximum-minimum normalization and Z-score normalization. After correction and standardization, the server extracts two key metrics: Skin Conductance Level (SCL) and Skin Conductance Response (SCR). SCL reflects the user's overall arousal level, while SCR reflects the intensity of the user's transient response to a specific stimulus.

[0051] Finally, the server combines the PSD features of the brainwave rhythm, the RMS value of electromyography, and the SCL and SCR of skin conductance into a high-dimensional vector, which serves as the user's physiological feature vector in the current state. This vector integrates the responses of multiple physiological systems, including the central nervous system, peripheral nervous system, and autonomic nervous system, and can comprehensively depict the user's physiological and emotional state.

[0052] For example, when a user views an image of a character that evokes positive emotions, the following physiological responses may occur: increased alpha wave energy and decreased beta wave energy on the electroencephalogram (EEG); increased facial electromyography (EMG) amplitude, resulting in a smiling expression; and increased skin conductance levels, showing a distinct skin conductance response peak. The server extracts these physiological indicators to construct a physiological feature vector representing a state of pleasure and relaxation. Conversely, if a user views an image of a disturbing character, responses such as decreased alpha waves, increased beta waves, decreased EMG amplitude, and decreased skin conductance levels may occur, with the physiological feature vector representing a state of tension and anxiety.

[0053] Step S104: Based on the preset character attribute space, generate a candidate character pool containing multiple candidate characters, wherein a candidate character is determined by a set of candidate parameters.

[0054] In step S104, the server defines a character attribute space, which contains various attribute dimensions that constitute a character, such as gender, age, height, weight, hairstyle, skin color, personality, and skills. Each attribute dimension has its own value range and type; for example, gender is usually binary (male / female), age is a continuous numerical value, and personality is a discrete category (such as cheerful, introverted, brave, etc.). Then, based on the character attribute space, the server generates a set of candidate characters through methods such as random sampling or heuristic search. Each candidate character corresponds to an attribute vector, and each element of the attribute vector represents the value of the character in a certain attribute dimension. When generating candidate characters, the correlation constraints between attributes are considered, such as the positive correlation between age and height, and the matching between gender and clothing style. These constraints can be obtained through expert knowledge, statistical analysis, etc., and embedded into the candidate character generation algorithm.

[0055] For example, suppose the character attribute space includes the following dimensions: gender (male / female), age (0-100), height (150-200cm), profession (warrior, mage, archer, etc.), and personality (cheerful, introverted, neutral). The server can generate a candidate character through the following steps: randomly sample gender, such as drawing "male"; sample age based on gender, such as a male character's age following a normal distribution with a mean of 25 and a standard deviation of 5, resulting in an age of 28; sample height based on gender and age, such as a 28-year-old male character's height following a normal distribution with a mean of 175cm and a standard deviation of 10cm, resulting in a height of 180cm; randomly sample profession, such as drawing "warrior"; sample personality based on gender, age, and profession, such as a young male warrior's personality being predominantly cheerful, resulting in "cheerful". After the above steps, the attribute vector of a candidate character is obtained as: [male, 28 years old, 180cm, warrior, cheerful]. Repeat this process to generate multiple different candidate characters, forming a candidate character pool.

[0056] Step S105: Map the player preference model to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character.

[0057] In step S105, the player preference model is mapped to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character. Specifically, this includes: mapping the player preference model to each dimension of the preset character attribute space to obtain the user's preference distribution in each attribute dimension; for each attribute dimension, the expected value of the attribute dimension is calculated according to the preference distribution, and the expected value is used as the target parameter of the ideal character in the corresponding attribute dimension.

[0058] Specifically, the server maps the player preference model to a pre-defined character attribute space to obtain the player's ideal character and the target parameter values ​​for that character across various attribute dimensions. First, the server maps the player preference model to the various dimensions of the character attribute space. The player preference model is a high-dimensional vector, where each dimension represents the player's degree of liking for a certain character element, such as a preference for male characters or magic classes. The character attribute space, on the other hand, is a structured feature space, where each dimension represents a specific attribute of the character, such as gender, age, height, or occupation. Mapping the preference model to the attribute space involves finding the correspondence between the two spaces, associating each element in the preference vector with one or more dimensions in the attribute space.

[0059] Specifically, the server can use the following mapping methods: First, one-to-one mapping, which directly maps each element in the preference vector to a dimension in the attribute space. For example, mapping "preference for male characters" to the "gender" dimension, and "preference for magic professions" to the "profession" dimension. This mapping method is simple and intuitive, but may overlook the correlation between different preferences. Second, one-to-many mapping: mapping each element in the preference vector to multiple dimensions in the attribute space. For example, mapping "preference for cute styles" to multiple dimensions such as "age," "personality," and "clothing," indicating that players like young, lively characters with cute outfits. This mapping method can capture the complex relationships between preferences, but requires careful design of the mapping rules.

[0060] Second, there is the many-to-many mapping, which combines multiple elements in the preference vector and maps them to multiple dimensions in the attribute space. For example, combining "preference for male characters" and "preference for magic professions" maps to multiple dimensions such as "gender," "profession," and "skills," indicating that players prefer male magician characters. This mapping method is more flexible and can uncover the implicit relationships between preference combinations and attribute combinations.

[0061] Next, the server calculates the expected value for each attribute dimension based on the preference distribution, which serves as the target parameter for the ideal character in that attribute. The method for calculating the expected value depends on the type of attribute and the form of the preference distribution. Common methods include: for discrete attributes, selecting the value with the highest preference probability as the expected value; for example, in the "gender" dimension, if the preference probability of "male" is 0.8, then "male" is used as the target parameter. For continuous attributes, the expected value is calculated based on statistics such as the mean and median of the preference distribution; for example, in the "age" dimension, if the mean of the preference distribution is 25, then 25 years old is used as the target parameter. For ordered discrete attributes, the expected value is calculated based on the preference distribution and the order of attribute values; for example, in the "body shape" dimension, if the preference distribution is "thin: 0.2, medium: 0.6, fat: 0.2", then "medium" is used as the target parameter.

[0062] Step S106: Based on the target parameters and the candidate parameters corresponding to each candidate role, determine the actual role in the candidate role pool and display the actual role to the user.

[0063] In step S106, the step of determining the actual role in the candidate role pool based on the target parameter and the candidate parameters corresponding to each candidate role, and displaying the actual role to the user, specifically includes: for each candidate role in the candidate role pool, extracting the corresponding candidate parameters; calculating the similarity between the candidate parameter vector and the target parameter vector; sorting the candidate roles according to the similarity, obtaining the candidate role with the highest similarity to the target parameter as the actual role, and displaying the actual role to the user.

[0064] Specifically, the server selects the candidate character that best matches the target parameters from the candidate character pool and recommends it to the player as the actual character. First, the server extracts the corresponding candidate parameters for each candidate character in the pool. A candidate parameter is a multi-dimensional vector, where each dimension represents the character's value for a certain attribute, such as gender, age, height, or occupation. The method for extracting candidate parameters depends on the representation of the candidate character. If the candidate character is stored as structured data (e.g., attribute-value pairs), the value of each attribute can be directly read to form the candidate parameter vector. If the candidate character is stored as unstructured data (e.g., images or text descriptions), feature extraction is performed first. For example, computer vision algorithms are used to extract features such as gender and age from images, and natural language processing algorithms are used to extract features such as occupation and personality from text. The extracted feature values ​​are then combined to form the candidate parameter vector.

[0065] The process of extracting candidate parameters can be pre-completed when generating the candidate role pool to improve the efficiency of real-time matching. Next, the server calculates the similarity between the candidate parameter vector and the target parameter vector for each candidate role. Similarity measures how close two vectors are in the attribute space; similarity metrics include Euclidean distance, cosine similarity, and Jaccard similarity. Different similarity metrics are suitable for different types of attributes and data distributions; for example, Euclidean distance is suitable for continuous attributes, while Jaccard similarity is suitable for discrete attributes. The server can choose an appropriate similarity metric based on specific circumstances, or it can combine multiple metrics.

[0066] After calculating the similarity of all candidate characters, the server sorts them according to similarity, resulting in a list of candidate characters ranked from highest to lowest according to their match with the target parameters. The candidate character at the top of the list has parameters closest to the target parameters, indicating that it best matches the player's preferences and expectations. The server can directly select the candidate character with the highest similarity as the actual character. Finally, the server displays the selected actual character to the player. This can be done by generating a 3D model of the character in the game interface, allowing the player to try it out and interact with it; or by displaying the character's attributes, backstory, special abilities, and other information in the character selection interface for the player's reference and selection.

[0067] Reference Figure 2This application also provides a personalized virtual character creation device, which is a server. The server includes an acquisition module 201 and a processing module 202, wherein: the acquisition module 201 is used to acquire visual and auditory stimulus materials reflecting various character elements and display the visual and auditory stimulus materials to the user; the acquisition module 201 is also used to acquire the user's physiological data for various character elements, including electroencephalogram (EEG) data, electromuscular wave data, and electrodermal response data; the processing module 202 is used to analyze the physiological data to obtain the user's preference features for various character elements and quantify the preference features into points in a high-dimensional vector space to form a player preference model; the processing module 202 is also used to generate a candidate character pool containing multiple candidate characters based on a preset character attribute space, wherein a candidate character is determined by a set of candidate parameters; the processing module 202 is also used to map the player preference model to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character; the processing module 202 is also used to determine the actual character in the candidate character pool based on the target parameters and the candidate parameters corresponding to each candidate character, and display the actual character to the user.

[0068] In one possible implementation, the processing module 202 analyzes the physiological data to obtain the user's preference features for each role element. Specifically, the processing module 202 constructs a physiological feature vector based on the physiological data; the processing module 202 inputs the physiological feature vector into a preset machine learning model and obtains the score corresponding to each role element through inference by the preset machine learning model; the processing module 202 sorts each role element according to the score and obtains a preset number of target role elements as preference features.

[0069] In one possible implementation, the processing module 202 constructs a physiological feature vector based on physiological data, specifically including: preprocessing the EEG data and extracting the EEG rhythm from the EEG data using a bandpass filter; calculating the power spectral density from the EEG rhythm to obtain the power spectral density features of each frequency band; performing full-wave rectification and smoothing on the muscle electroencephalogram data to extract the root mean square value of muscle contraction; performing baseline correction and standardization on the skin conductance response data to obtain the skin conductance level and skin conductance response amplitude; and constructing a physiological feature vector based on the power spectral density features, root mean square value, skin conductance level, and skin conductance response amplitude.

[0070] In one possible implementation, the processing module 202 quantizes the preference features into points in a high-dimensional vector space to form a player preference model. Specifically, the processing module 202 constructs a character element dictionary and numbers each character element in the dictionary according to a preset order, obtaining a mapping relationship between the element ID and element name of each character element; the processing module 202 extracts the target character elements contained in each preference feature to obtain a list of element IDs, which includes multiple target element IDs, each target element ID being the element ID corresponding to a target character element; the processing module 202 creates a binary vector based on the length of the element ID list, the length of which is equal to the size of the character element dictionary, and the binary vector is an all-zero vector; the processing module 202 iterates through each target element ID in the element ID list, determines the corresponding position of each target element ID in the binary vector, and sets the value of the corresponding position to 1, obtaining a preference vector corresponding to the preference feature; the preference vector is mapped to a preset high-dimensional vector space, and a probability distribution estimate is performed on the mapped preference vector to obtain the player preference model.

[0071] In one possible implementation, the processing module 202 generates a candidate character pool containing multiple candidate characters based on a preset character attribute space. Specifically, the processing module 202 defines a preset character attribute space, which includes an appearance attribute dimension, a personality attribute dimension, and a skill attribute dimension; the processing module 202 determines the attribute categories included in each of the appearance attribute dimension, personality attribute dimension, and skill attribute dimension; the processing module 202 determines the candidate attribute parameters corresponding to each attribute category; and the processing module 202 combines the candidate attribute parameters in each attribute category to generate candidate parameters corresponding to multiple candidate characters, thus obtaining the candidate character pool.

[0072] In one possible implementation, the processing module 202 maps the player preference model to a preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character. Specifically, the processing module 202 maps the player preference model to each dimension of the preset character attribute space to obtain the user's preference distribution in each attribute dimension. For each attribute dimension, the processing module 202 calculates the expected value of the attribute dimension according to the preference distribution and uses the expected value as the target parameter of the ideal character in the corresponding attribute dimension.

[0073] In one possible implementation, the processing module 202 determines the actual role in the candidate role pool based on the target parameter and the candidate parameters corresponding to each candidate role, and displays the actual role to the user. Specifically, the processing module 202 extracts the corresponding candidate parameters for each candidate role in the candidate role pool; the processing module 202 calculates the similarity between the candidate parameter vector and the target parameter vector; the processing module 202 sorts the candidate roles according to the similarity, obtains the candidate role with the highest similarity to the target parameter as the actual role, and displays the actual role to the user.

[0074] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0075] This application also provides an electronic device. (See reference...) Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0076] The communication bus 302 is used to enable communication between these components.

[0077] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0078] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0079] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0080] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for creating personalized virtual characters.

[0081] exist Figure 3In the illustrated electronic device 300, the user interface 303 is mainly used to provide an input interface for the user and to acquire user input data; while the processor 301 can be used to call an application stored in the memory 305 for a personalized virtual character creation method. When executed by one or more processors 301, the electronic device 300 performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0082] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors 301, these instructions cause an electronic device 300 to perform one or more of the methods described in the above embodiments.

[0083] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0084] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0085] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0086] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0087] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0088] The above description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the specification and the disclosure of practical truths.

[0089] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for creating personalized virtual characters, characterized in that, The method includes: Acquire visual and auditory stimuli that reflect the elements of each character, and display the visual and auditory stimuli to the user; Acquire the user's physiological data for each of the aforementioned character elements, including electroencephalogram (EEG) data, electromuscular wave (EMG) data, and electrodermal response (EDR) data. The physiological data is analyzed to obtain the user's preference features for each of the character elements, and the preference features are quantified into points in a high-dimensional vector space to form a player preference model; Based on a preset character attribute space, a candidate character pool containing multiple candidate characters is generated, wherein each candidate character is determined by a set of candidate parameters; The player preference model is mapped to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character; Based on the target parameters and the candidate parameters corresponding to each candidate role, the actual role is determined in the candidate role pool and the actual role is displayed to the user; The analysis of the physiological data to obtain the user's preference characteristics for each of the role elements specifically includes: Construct a physiological feature vector based on the physiological data; The physiological feature vector is input into a preset machine learning model, and the score corresponding to each of the role elements is obtained by reasoning through the preset machine learning model. The character elements are sorted according to the scores to obtain a preset number of target character elements as the preference features; The construction of physiological feature vectors based on the physiological data specifically includes: The EEG data is preprocessed, and the EEG rhythm is extracted from the EEG data using a bandpass filter; The power spectral density is calculated from the brainwave rhythms to obtain the power spectral density characteristics of each frequency band; The muscle electroencephalogram data is subjected to full-wave rectification and smoothing to extract the root mean square value of muscle contraction. The skin conductance response data were baseline-corrected and standardized to obtain the skin conductance level and skin conductance response amplitude. A physiological feature vector is constructed based on the power spectral density characteristics, the root mean square value, the skin conductance level, and the skin conductance response amplitude. The step of quantizing the preference features into points in a high-dimensional vector space to form a player preference model specifically includes: Construct a character element dictionary, and number each character element in the character element dictionary according to a preset order to obtain the mapping relationship between the element ID and the element name of each character element; Extract the target role element contained in each of the preference features to obtain an element ID list. The element ID list includes multiple target element IDs, and the target element ID is the element ID corresponding to the target role element. Based on the length of the element ID list, a binary vector is created. The length of the binary vector is equal to the size of the character element dictionary. The binary vector is an all-zero vector. Traverse each target element ID in the element ID list, determine the corresponding position of each target element ID in the binary vector, and set the value of the corresponding position to 1 to obtain the preference vector corresponding to the preference feature; The preference vector is mapped to a preset high-dimensional vector space, and the probability distribution of the mapped preference vector is estimated to obtain the player preference model.

2. The method according to claim 1, characterized in that, The process of generating a candidate character pool containing multiple candidate characters based on a preset character attribute space specifically includes: Define the preset character attribute space, which includes appearance attribute dimension, personality attribute dimension and skill attribute dimension; In the appearance attribute dimension, the personality attribute dimension, and the skill attribute dimension, the attribute categories included in each attribute dimension are determined respectively; For each of the attribute categories, candidate attribute parameters corresponding to each attribute category are determined; The candidate attribute parameters in each attribute category are combined to generate candidate parameters corresponding to multiple candidate roles, thus obtaining the candidate role pool.

3. The method according to claim 1, characterized in that, The step of mapping the player preference model to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character specifically includes: The player preference model is mapped to each dimension of the preset character attribute space to obtain the user's preference distribution in each attribute dimension; For each attribute dimension, the expected value of the attribute dimension is calculated according to the preference distribution, and the expected value is used as the target parameter of the ideal character in the corresponding attribute dimension.

4. The method according to claim 1, characterized in that, The step of determining the actual role from the candidate role pool based on the target parameters and the candidate parameters corresponding to each candidate role, and displaying the actual role to the user, specifically includes: For each candidate character in the candidate character pool, extract the corresponding candidate parameters; Calculate the similarity between the candidate parameter vector and the target parameter vector; The candidate roles are sorted according to the similarity, and the candidate role with the highest similarity to the target parameter is selected as the actual role, which is then displayed to the user.

5. A personalized virtual character creation device, characterized in that, The apparatus is used to perform the method as described in any one of claims 1-4, the apparatus comprising an acquisition module (201) and a processing module (202), wherein: The acquisition module (201) is used to acquire visual and auditory stimulus materials reflecting each character element, and to display the visual and auditory stimulus materials to the user; The acquisition module (201) is also used to acquire the user's physiological data for each of the role elements, the physiological data including brainwave data, muscle wave data and skin conductance data; The processing module (202) is used to analyze the physiological data, obtain the user's preference features for each of the role elements, and quantify the preference features into points in a high-dimensional vector space to form a player preference model; The processing module (202) is also used to generate a candidate character pool containing multiple candidate characters based on a preset character attribute space, wherein a candidate character is determined by a set of candidate parameters; The processing module (202) is also used to map the player preference model to the preset character attribute space to obtain the ideal character and the target parameters corresponding to the ideal character; The processing module (202) is further configured to determine the actual role in the candidate role pool based on the target parameters and the candidate parameters corresponding to each candidate role, and display the actual role to the user.

6. An electronic device, characterized in that, The device includes a processor (301), a memory (305), a user interface (303), and a network interface (304). The memory (305) is used to store instructions. The user interface (303) and the network interface (304) are used to communicate with other devices. The processor (301) is used to execute the instructions stored in the memory (305) to cause the electronic device (300) to perform the method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-4.

Images