Cross-account game resource intelligent loading method and system based on biometric recognition

Abstract

This invention relates to a cross-account intelligent loading method and system for game resources based on biometric recognition, belonging to the field of data processing technology. It includes collecting player biometric data, player behavior data, and spatial heatmaps; obtaining a player's biometric ID by recognizing the player's biometric data; combining the player's behavior data and spatial heatmaps to obtain a player behavior profile, and binding it with the player's biometric ID to obtain a biometric behavior profile library; collecting real-time player biometric data; extracting the real-time player's biometric ID when it is inconsistent with the player's biometric data; retrieving the real-time player's biometric ID and matching it against the biometric behavior profile library to obtain a real-time player behavior profile; and executing a preset priority loading method for game resources based on the real-time player behavior profile to obtain a priority loading sequence and an idle asynchronous loading queue, and then loading the game resources. This invention improves resource loading efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for intelligent loading of cross-account game resources based on biometric recognition. Background Technology

[0002] Game resource loading technology is a technique designed to ensure a smooth gaming experience by efficiently managing the loading process of these resources.

[0003] Currently, game resource loading is mainly based on a static resource list verification loading mechanism, modular management of resource packages, differential update technology that generates difference files between old and new versions through binary differential algorithms, and lazy loading and asynchronous loading methods.

[0004] The aforementioned technologies generally employ account-based behavioral analysis. However, if a player registers a new account, switches to a secondary account, or logs in on another platform, game resources cannot be accurately loaded for that player. This results in a large amount of resource loading, leading to delays and lag, and reducing resource loading efficiency. Summary of the Invention

[0005] To improve resource loading efficiency, this invention provides a cross-account game resource intelligent loading method and system based on biometric recognition.

[0006] In a first aspect, the present invention provides a cross-account game resource intelligent loading method based on biometric recognition, employing the following technical solution: A cross-account game resource intelligent loading method based on biometric recognition, comprising: Collect player biometric data, player behavior data, and spatial heat vectors; Player biometric IDs are obtained by identifying player biometric data; Player behavior data and spatial heat vectors are input into a preset conditional probability model to obtain a player behavior profile; Upload player biometric IDs and player behavioral profiles to the cloud server and bind them to obtain a biometric behavioral profile database; Collect real-time player biometric data during account login; When real-time player biometric data is inconsistent with player biometric data, extract the real-time player biometric ID based on the real-time player biometric data; Real-time player biometrics are obtained by retrieving and matching real-time player biometrics with a biometrics behavior profile database. Based on real-time player behavior profiles, the game resources are loaded using a pre-defined priority loading method to obtain a priority loading sequence and an idle asynchronous loading queue, and then the game resources are loaded.

[0007] By adopting the above technical solution, player biometric data is collected to generate a unique biometric ID, and player behavior profiles are obtained by combining player behavior data with spatial heat vectors, thus forming a biometric behavior profile library. When a player logs in under different accounts, the current operator's identity is identified by comparing real-time biometric data, and the unified behavior profile across accounts is retrieved. The priority loading method is executed according to the profile, thereby reducing the resource waiting time when players switch between different accounts and improving resource loading efficiency.

[0008] Optionally, methods for acquiring spatial thermal vectors are also included: Collect data from each content module; Each content module is equally divided into multiple grid units to obtain the grid units for each module; Collect information on the duration of player stay in each module's grid cell; The thermal values ​​of grid cells are extracted based on the dwell time information; The spatial thermal vector is obtained by performing ensemble analysis on the thermal values ​​of the grid cells.

[0009] By using the above technical solution, the game's various content modules are equally divided into multiple grid units. The player's dwell time information in each module's grid unit is collected, the grid unit heat value is extracted and analyzed in sets to generate a spatial heat vector, which improves the accuracy of profile construction and the targeting of loading strategies.

[0010] Optionally, methods for establishing conditional probability models are also included: Collect historical behavioral data from player accounts; The state transition frequency requirement is obtained by calculating the state transition frequency based on historical behavior data. The conditional probability model is trained by the state transition frequency requirement; Collect players' real-time behavior data; Compare players' real-time behavior data with their historical behavior data to obtain player behavior difference values ​​and accumulate them; When the accumulated player behavior difference value exceeds the preset behavior difference threshold, the state transition frequency requirement is collected and updated again, and the parameters of the conditional probability model are updated based on the updated state transition frequency requirement.

[0011] By using the above technical solution, historical behavior data of player accounts on the server side is collected, state transition frequency is calculated and conditional probability models are trained. At the same time, real-time behavior data of players is continuously collected and compared with historical behavior data to obtain behavior difference values. When the cumulative difference value exceeds a preset threshold, the model parameters are updated, which improves the long-term accuracy and reliability of resource preloading.

[0012] Optionally, a cold start loading method for new players with no historical behavior data is also included: When no real-time player behavior profile corresponding to the real-time player's biological ID is found in the biological behavior profile database, the player is determined to be a new player, and the game information stored locally on the terminal device used by the new player is collected. Extract local behavior data from game information stored locally; Extract behavioral preference features from local behavioral data; An initial behavioral profile of the new player is generated based on behavioral preference characteristics. This initial behavioral profile is then bound to the real-time player's biometric ID and uploaded to the biometric profile library on the cloud server. Game resources are then loaded with priority based on the initial behavioral profile.

[0013] Through the above technical solution, for new players without historical behavior data, game information stored locally on the new player's terminal device is collected, behavioral preference features are extracted from it and an initial behavioral profile is generated. After binding it with the biometric ID, it is uploaded to the cloud, which further improves the resource loading speed.

[0014] Optionally, the priority loading method may also be included: Collect information about the resource packages that need to be loaded; Identify player behavior preferences based on real-time player behavior profiles; By extracting the information of the resource packs that need to be loaded from the player's behavioral preference information, candidate resource packs and non-candidate resource packs are obtained; Extract the frequency and timing of candidate resource packs in player behavior preference information; The priority weight value is obtained by weighting the candidate resource packs based on their frequency and time of appearance in the player's behavioral preference information. All candidate resource packages are sorted from highest to lowest priority weight to obtain a priority loading queue, and priority loading is performed according to the priority loading queue. Non-candidate resource packages are uniformly arranged into an idle asynchronous loading queue.

[0015] By using the above technical solution, information on resource packages that need to be loaded is collected, and behavioral preference information is confirmed by combining real-time player behavior profiles. Candidate resource packages are extracted and weighted according to their frequency and time of appearance in the behavioral preference information to form a priority loading queue. The resource loading order is precisely matched with the player's personal preferences, thus optimizing the overall utilization efficiency of network resources.

[0016] Optionally, the collection of game information locally stored in the terminal device used by the new player specifically includes: Request the terminal device's operating system to read the database file from the local storage space and collect the request result; If the application result is "authorization granted", the database file is scanned and candidate game information is obtained by identifying the game type classification table. Retrieve game log information based on candidate game information; Game log information is extracted to obtain locally stored game information.

[0017] The above technical solution requests the terminal device's operating system to scan database files and identifies candidate game information through a game type classification table. It retrieves game log information and extracts locally stored game information, enabling the system to automatically discover and extract the closest type of game behavior data already existing in the terminal device, providing a reliable data source for generating initial profiles of new players.

[0018] Optionally, it also includes methods for processing profiles of players who have been inactive for a long time: Iterate through the bio profile records corresponding to each bio ID in the bio behavior profile database and extract the last active timestamp of each bio profile record; Collect the current system time and calculate the difference between the current system time and the last active timestamp to obtain the inactivity duration of the profile; When the inactive time exceeds the preset quit time threshold, the biometric ID is marked as a quitting player, and the corresponding player behavior profile is marked as a profile to be compressed. The compressed image is compressed and then stored back into the biological behavior image database. When a player who has been marked as a quitter is detected to log in again via biometric identification, the player's profile is decompressed and the real-time updates of that profile are restored.

[0019] By using the above technical solution, the last active timestamp of each image is extracted from the bio-behavioral profile database, the inactivity time is calculated and compared with the preset quit time threshold, and the quit player profiles that exceed the threshold are marked as profiles to be compressed and compression is performed. This effectively manages the storage space of the cloud bio-behavioral profile database and balances storage efficiency and service quality.

[0020] Optionally, a smart resource loading method for returning players is also included: When a player who has been marked as a quitter is detected to log in again via biometric identification, that player is marked as a returning player. Collect the biometric data of returning players to obtain their IDs; The behavioral profiles of returning players are obtained by traversing and searching the biometric profile database based on the returning player IDs. Collect promotional information from the game's official website within a preset time period and keywords related to discussions about the game on mainstream social networking sites, and compare them with returning player behavior profiles to obtain player return keywords; Based on player return keywords, the resource package information to be loaded is extracted to obtain player return resource information, and the game resources are loaded according to priority based on the player return resource information.

[0021] The above technical solution detects that a player who has quit the game has logged in again, and marks them as a returning player. Based on the returning player's ID, the corresponding behavioral profile is retrieved. At the same time, promotional information from the game's official website and discussion keywords from mainstream social networking sites within a preset time period are collected. Based on this, the player's returning resource information is extracted, and a priority loading method is used to load game resources, thereby improving the accuracy of game resource loading.

[0022] Optionally, an iterative update method for the initial behavioral profile may also be included: Collect the number of game sessions for new players; When the number of game sessions for a new player exceeds the preset threshold, collect the personal behavior sequence data generated by the new player during the game session; Input individual behavioral sequence data into a conditional probability model to generate a unique behavioral profile for new players; Calculate the difference between the customized behavioral profile and the initial behavioral profile; When the difference exceeds the preset profile difference threshold, the initial behavioral profile is replaced with a unique behavioral profile, and the unique behavioral profile is re-bound to the new player's biometric ID before being uploaded to the biometric behavioral profile library on the cloud server.

[0023] By adopting the above technical solution, the number of game sessions of new players is collected. When the number of sessions exceeds a preset threshold, their personal behavior sequence data is collected and input into a conditional probability model to generate a unique behavior profile. The difference between the unique profile and the initial profile is calculated. The unique profile replaces the initial profile and is re-bound to the cloud, which improves the accuracy of the profile and the effectiveness of subsequent loading strategies.

[0024] Secondly, this application provides a cross-account game resource intelligent loading system based on biometric recognition, employing the following technical solution: A cross-account game resource intelligent loading system based on biometric recognition, comprising: The data acquisition module is used to collect player biometric data, player behavior data, spatial heat vectors, and real-time player biometric data. The memory is used to store the program that implements a cross-account game resource intelligent loading method based on biometric recognition; The processor is used to load and execute programs stored in memory.

[0025] In summary, this application includes at least one of the following beneficial technical effects: 1. Collect player biometric data to generate a unique biometric ID, and combine player behavior data with spatial heat vectors to obtain player behavior profiles, thereby forming a biometric behavior profile library; when a player logs in under different accounts, compare real-time biometric data to identify the current operator's identity, retrieve their unified behavior profile across accounts, and execute priority loading methods based on the profile, thereby reducing resource waiting time when players switch between different accounts and improving resource loading efficiency. 2. For new players without historical behavior data, game information stored locally on the new player's terminal device is collected, behavioral preference features are extracted from it and an initial behavioral profile is generated. After being bound to the biometric ID, it is uploaded to the cloud, thus optimizing resource loading speed for new players; 3. When a player who has quit the game logs back in, they are marked as a returning player. Based on the returning player's ID, their corresponding behavioral profile is retrieved. At the same time, promotional information from the game's official website and discussion keywords from mainstream social media sites within a preset time period are collected. Based on this, the player's returning resource information is extracted, and a priority loading method is used to load game resources, thereby improving the accuracy of game resource loading. Attached Figure Description

[0026] Figure 1 This is a flowchart of a cross-account game resource intelligent loading method based on biometric recognition; Figure 2 This is a flowchart illustrating the specific method for intelligent loading of game resources for new players; Figure 3 This is a flowchart illustrating the specific method for intelligently loading game resources for returning players. Detailed Implementation

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0028] Reference Figure 1 , Figure 2 and Figure 3 This application discloses a method for intelligent loading of cross-account game resources based on biometric recognition, including the following steps: S10: Collect player biometric data, player behavior data, and spatial heat vectors.

[0029] Player biometric data refers to the data contained in a player's biometric features, such as facial data, voiceprint data, and fingerprint data.

[0030] Player behavior data refers to the data on player actions during gameplay, including but not limited to: login time, game mode selection, character preference, and frequency of match participation.

[0031] Spatial heat vector refers to the vector formed by quantifying the player's dwell time and operation frequency in various content modules of the game, representing the heat value of each grid unit. The specific collection method is described in detail in S20 to S24, and will not be repeated here.

[0032] The grid cell heat value is a quantified value obtained by quantifying the time a player spends and the frequency of their actions in various content modules of the game.

[0033] After the player grants biometric data collection permission, the player's biometric data is collected through the fingerprint sensor or facial feature collector built into the terminal device. The player's behavior data generated during the game is collected through the behavior log recording module on the client and the spatial heat vector is obtained by mapping the grid cell heat value. The specific mapping method is described in detail in S24 and will not be repeated here.

[0034] S11: Obtain the player's biometric ID by recognizing the player's biometric data.

[0035] A player's biometric ID is a unique ID used to distinguish a player's identity.

[0036] To protect the privacy of biometric data, a one-way hash operation is performed on the player's biometric data, which is stored only on the local client, to generate a unique identifier that distinguishes different players, i.e., the player's biometric ID.

[0037] S12: Input player behavior data and spatial heat vectors into a preset conditional probability model to obtain a player behavior profile.

[0038] A conditional probability model is a state transition model based on Markov chains, constructed using the frequency of state transitions, used to predict the player's next action. This prediction model is established in advance by those skilled in the art, and the specific methods are described in detail in S30 to S35, and will not be repeated here.

[0039] State transition frequency refers to the cumulative number of times a player transitions from one game state to another in the player's historical behavior data.

[0040] Historical behavior data refers to the collection of a player's operational behavior records in various game scenarios, which are recorded and stored by the server during the player's past gameplay.

[0041] A player behavior profile is a set of features used to describe a player's gaming habits and content preferences.

[0042] The state transition frequency is extracted from player behavior data. The player behavior data contains the state transition frequency. The state transition frequency and the spatial heat vector are concatenated into a unified trend vector as input parameters. After the input parameters are transformed by the conditional probability model, the output is the player behavior profile.

[0043] S13: Upload the player's bio-ID and player behavior profile to the cloud server and bind them to obtain the bio-behavioral profile library.

[0044] The android behavior profile database refers to a database deployed on a cloud server that stores the binding relationship between each player's android ID and the corresponding player behavior profile.

[0045] By uploading and binding the player's biometric ID and player behavior profile to the cloud server through the network communication module of the player's terminal device, a biometric behavior profile database is obtained, which can centrally store and facilitate the retrieval of the player behavior profile corresponding to the player's biometric ID when crossing accounts.

[0046] S14: Collect real-time player biometric data during account login.

[0047] Real-time player biometric data refers to the biometric information of a player collected in real time by the biometric sensor of the terminal device when the player triggers the account login operation.

[0048] Biometric sensors refer to fingerprint sensors or facial feature collectors built into terminal devices.

[0049] S15: When the real-time player biometric data is inconsistent with the player biometric data, extract the real-time player biometric ID based on the real-time player biometric data.

[0050] Real-time player biometric ID refers to the player's biometric information collected in real time by the biometric sensor of the terminal device when the player triggers the account login operation, and the corresponding player biometric ID.

[0051] When a discrepancy is detected between the real-time player biometric data and the player's biometric data, a one-way hash operation is used to process the real-time player biometric data and extract the real-time player's biometric ID, thereby identifying the true identity of the current real-time operator.

[0052] S16: Retrieve real-time player bio IDs and match them with the bio behavior profile database to obtain real-time player behavior profiles.

[0053] Real-time player behavior profiles refer to player behavior profiles based on real-time player biometric IDs.

[0054] The database query engine iterates and matches real-time player avatar IDs with the avatar behavior profile database to obtain real-time player behavior profiles.

[0055] S17: Based on the real-time player behavior profile, execute the game resource preset priority loading method to obtain the priority loading sequence and idle asynchronous loading queue, and load the game resources.

[0056] Priority loading sequence refers to a loading queue that predicts the resource packs that a player is likely to use in the current game session based on the player's behavior profile and arranges them in descending order of priority. Resources in this queue are downloaded and loaded first when the game starts or when the background is idle.

[0057] The idle-time asynchronous loading queue refers to a loading queue consisting of resource packages that are predicted to be used with low probability in the current game session. These packages are downloaded gradually in an asynchronous manner with low priority during system idle periods.

[0058] By executing a preset priority loading method through a resource loading scheduler, a priority loading sequence and an idle asynchronous loading queue are generated. This achieves the effect of prioritizing the loading of resources with a high probability of use and loading low-probability resources asynchronously in the background, thereby reducing the waiting time for players to switch scenes.

[0059] Priority loading method refers to the method of predicting resource usage probability based on player behavior profile, and then sorting the resources to be loaded from high to low according to the resource usage probability and loading them at different speeds. This will be explained in detail in S50 to S55, and will not be repeated here.

[0060] It also includes methods for acquiring spatial thermal vectors: S20: Collect data from each content module.

[0061] Each content module refers to a different section obtained after dividing the game content.

[0062] The game engine editor divides all content data of the game into different data function types, such as the store content module, the combat content module, and the home content module. The specifics are set in advance by those skilled in the art, thus obtaining each content module.

[0063] S21: Divide each content module equally into multiple grid units to obtain the grid units of each module.

[0064] Each module grid unit refers to the smallest unit that can be quantified and statistically analyzed by discretizing the interface space of each content in the game. Its size is set in advance by those skilled in the art.

[0065] By discretizing the continuous game interface space in each content module of the game into quantifiable and statistically significant grid units, every click, hover, and stay of the player on the interface can be accurately mapped to specific grid coordinates, which facilitates the subsequent statistical analysis of the heat value of each spatial location.

[0066] S22: Collect information on the duration of player stay in each module grid cell.

[0067] The dwell time information refers to the total dwell time of the player when they click on the content corresponding to that cell.

[0068] The client behavior log recording module collects information on the duration of time players spend in the content corresponding to each module grid unit.

[0069] S23: Extract the thermal value of the grid cell based on the dwell time information.

[0070] The grid cell heat value refers to the specific value after quantifying the player's preferences.

[0071] The thermal value of the grid cell is obtained by weighting and extracting the residence time information.

[0072] The weighted extraction process includes counting the frequency of players entering each grid cell. For grid cells with high-frequency visits but short stays, the stay duration of each grid cell is weighted upward according to the frequency. For example, a stay of 20 seconds multiplied by a frequency increase factor of 1.2 yields a weighted result of 24, which reflects the importance of the operation intention. A value weight value is preset for different grid cells. For example, the weight of the purchase button area in the store is greater than the weight value of the empty table display area. The stay duration of each grid cell is multiplied by a fixed frequency increase factor of 1 and then multiplied by the preset value weight to obtain the grid cell heat value.

[0073] Preset value weights refer to the weight values ​​that are set in advance by those skilled in the art based on different grid cells.

[0074] S24: Perform ensemble analysis on the thermal values ​​of the grid cells to obtain the spatial thermal vector.

[0075] By integrating the thermal values ​​of each grid cell, a spatial thermal vector is obtained, which can transform the player's attention to different game content dimensions into a computable multi-dimensional vector.

[0076] The specific integration steps are as follows: Arrange the heat values ​​of all grid units within a content module according to their spatial coordinates or row and column numbers to form a heat value matrix for that module. The position of each element in the matrix corresponds to the physical position of the grid on the interface, and the value is the heat value of that grid. The value reflects the player's attention to that area. After removing redundancy, the final multidimensional numerical sequence is the spatial heat vector.

[0077] It also includes methods for establishing conditional probability models: S30: Collect historical behavior data of player accounts.

[0078] Historical behavior data refers to the sum of all previous player behavior data.

[0079] Historical behavior data is obtained by retrieving and integrating all previous player behavior data from the server through the server's data filtering module.

[0080] S31: Calculate the state transition frequency requirement based on historical behavior data.

[0081] State transition frequency requirement refers to the statistical data set that needs to be extracted from behavioral data to reflect the frequency of transitions between various game states in order to accurately reflect the current behavior pattern of players when establishing or updating conditional probability models.

[0082] By using a state transition statistical algorithm to calculate the frequency of historical behavior data, the required state transition frequency is obtained, reflecting the switching patterns of players between different game states.

[0083] The state transition statistics algorithm is as follows: First, extract the game state corresponding to each behavior record from the historical behavior data. The game state is defined by the combination of the current scene area, the current UI interface type, and the most recent operation behavior type. Then, perform a sliding window scan on the behavior record sequence and construct state transition sample pairs from the preceding and following states within adjacent time windows. Count the transition frequency of all sample pairs to form a frequency matrix with the starting state as the row and the target state as the column. Finally, perform probability normalization on each row to obtain the conditional probability value of transitioning from each starting state to each target state. Use this probability distribution as the output of the state transition frequency requirement.

[0084] S32: Train the conditional probability model by the state transition frequency requirement.

[0085] Each behavior record corresponding to the state transition frequency requirement is input into the model as training material, enabling the model to learn the player's behavior pattern and predict the player's page switching in advance, thus obtaining a conditional probability model.

[0086] S33: Collect real-time behavioral data of players.

[0087] Real-time behavioral data refers to player behavior data at the current moment.

[0088] The same applies to S10, so I will not repeat it here.

[0089] S34: Compare the player's real-time behavior data with the player's historical behavior data to obtain the player behavior difference value and accumulate it.

[0090] Player behavior difference value refers to the difference between a player's real-time behavior data and the server's historical behavior data.

[0091] Real-time behavioral data is aggregated according to a preset time window to extract the real-time state transition frequency distribution. Then, the deviation of this real-time distribution from the historical state transition probability distribution constructed based on historical behavioral data is calculated item by item to obtain the deviation of each state transition item. The maximum deviation value is then selected as the player behavior difference value for a single comparison. Finally, the difference values ​​obtained from each comparison are accumulated. When the accumulated value exceeds the preset behavior difference threshold, the model is updated.

[0092] A preset time window refers to a fixed time value set in advance by those skilled in the art.

[0093] The behavioral difference threshold is a pre-established threshold set by those skilled in the art after multiple tests, used to quantify the maximum value of the normal range of variation in a player's behavioral profile.

[0094] S35: When the accumulated player behavior difference value exceeds the preset behavior difference threshold, collect and update the state transition frequency requirement again, and update the parameters of the conditional probability model based on the updated state transition frequency requirement.

[0095] When the cumulative behavioral difference value exceeds the behavioral difference threshold, the state transition frequency requirement needs to be updated, and the parameters of the conditional probability model need to be updated.

[0096] By incorporating real-time behavioral data into the historical behavioral data in S30, new historical behavioral data is formed. This is then used to perform step S31 again to update the state transition frequency requirement, thereby updating the parameters of the conditional probability model.

[0097] It also includes a cold start loading method for new players with no historical behavior data: S40: When no real-time player behavior profile corresponding to the real-time player's bio ID is found in the bio behavior profile database, the player is determined to be a new player, and the game information stored locally on the terminal device used by the new player is collected.

[0098] Cold start refers to the process of generating an initial behavioral profile for new players or players for whom there is no corresponding record in the behavioral profile database, and then loading resources accordingly, in the absence of personal historical behavioral data.

[0099] If no corresponding profile record is found in the biometric profile database, it indicates that the user is a new player. The game information stored locally on the terminal device used by the new player needs to be collected first for subsequent steps.

[0100] S41: Extract local behavior data from game information stored locally.

[0101] Local behavior data refers to player behavior data stored locally on the client side.

[0102] The client's resource manager retrieves locally stored game information to obtain local behavior data.

[0103] S42: Extract behavioral preference features from local behavioral data.

[0104] Behavioral preference features refer to a set of characteristic parameters that reflect a player's gaming habits and content preferences, extracted through analysis and statistical analysis from behavioral data of the player in this game or other similar games.

[0105] Behavioral preference features are obtained by extracting local behavioral data through game log management tools. The local behavioral data contains behavioral preference features.

[0106] S43: Generate an initial behavioral profile of the new player based on behavioral preference characteristics, bind the initial behavioral profile with the real-time player's biometric ID, upload it to the biometric profile library on the cloud server, and execute a priority loading method for game resources based on the initial behavioral profile.

[0107] The initial behavioral profile refers to the behavioral profile corresponding to a new player.

[0108] The local state transition frequency requirement is obtained by extracting behavioral preference features. Then, the local state transition frequency requirement is used to obtain an initial behavioral profile through a conditional probability model. This initial behavioral profile is then bound to the real-time player's biometric ID and uploaded to the biometric behavioral profile library. Finally, the resource loading scheduler executes a priority loading method based on the initial behavioral profile to achieve intelligent resource loading for new players as well.

[0109] The priority loading method includes: S50: Collect information about the resource packages that need to be loaded.

[0110] The resource package information to be loaded refers to the resource information that needs to be loaded in this game session.

[0111] The resource list management module collects the resource package information that needs to be loaded for this login.

[0112] S51: Confirm player behavior preference information based on real-time player behavior profiles.

[0113] Behavioral preference information refers to the specific set of data used to determine the behavioral habits of players.

[0114] Real-time player behavior profiles are retrieved to obtain behavioral preference information that guides resource loading priority determination. The real-time player behavior profiles contain behavioral preference information.

[0115] S52: Extract candidate and non-candidate resource packages from the resource package information to be loaded based on the player's behavioral preference information.

[0116] Candidate resource packs refer to the set of resource packs that are identified from the resource pack information that needs to be loaded, and are determined to be related to the player's behavioral preferences and have a high probability of being used in the current game session.

[0117] Non-candidate resource packs refer to the set of resource packs that need to be loaded, excluding candidate resource packs, and that have a low degree of matching or are irrelevant to player behavior and preference information.

[0118] The preference matching algorithm compares behavioral preference information with resource package information to obtain candidate resource packages that match the preferences and non-candidate resource packages that do not match.

[0119] The preference matching algorithm specifically involves: extracting a set of preference content identifiers from player behavior preference information and reading the associated content tags of each resource package; calculating the preference hit count as the matching degree by comparing the intersection of the set of preference content identifiers and the associated content tags of each resource package; marking resource packages with a matching degree reaching a preset matching threshold as candidate resource packages, and marking the rest as non-candidate resource packages.

[0120] The preference content identifier set refers to the set of key data in player behavior preference information.

[0121] Related content tags refer to tags in a resource package that represent data content.

[0122] S53: Extract the frequency and timing of candidate resource packs in player behavior preference information.

[0123] Frequency and timing refer to the number of times and the timing of cross-data between resource packs and player behavior preference information.

[0124] By comparing and retrieving candidate resource packs with players' behavioral preference information, the frequency and timing of the appearance of candidate resource packs in players' behavioral preference information can be obtained.

[0125] S54: The priority weight value is obtained by weighting the candidate resource packs based on their frequency and time of appearance in the player's behavioral preference information.

[0126] Priority weight value refers to the weight of candidate resource packs in the player's behavioral preference information.

[0127] The candidate resource packages are weighted based on their frequency of occurrence and most recent usage time using a weighted calculation formula to obtain priority weight values.

[0128] The weighted calculation formula comprehensively quantifies the candidate resource pack's appearance frequency and most recent usage time in player behavior preferences. First, the proportion of the candidate resource pack's appearances to the total appearances of all candidate resource packs is used as the normalized appearance frequency. Simultaneously, a time-decreasing weight value is calculated based on the time interval between the current time and the candidate resource pack's most recent usage; the longer the interval, the greater the decay. Then, the normalized appearance frequency and the time-decreasing weight are multiplied by their respective fixed coefficients and summed to obtain the candidate resource pack's priority weight value. The sum of the coefficients for the frequency and time dimensions is a fixed value, pre-set by those skilled in the art based on experience, to adjust the proportion of influence of the two dimensions on the final weight.

[0129] S55: Sort all candidate resource packages from high to low according to their priority weight values ​​to obtain a priority loading queue, and load them according to the priority loading queue. Non-candidate resource packages are uniformly arranged into an idle asynchronous loading queue.

[0130] All candidate resource packages are sorted from highest to lowest priority weight to obtain a priority loading queue. Downloads are prioritized according to the priority loading queue. Non-candidate resource packages are arranged into an idle asynchronous loading queue to ensure that high-frequency resources are available and low-frequency resources do not block the main thread.

[0131] The collection of game information locally stored on the terminal device used by the new player specifically includes: S60: Requests the terminal device's operating system to read the database file in the local storage space and collects the request result.

[0132] The application result refers to whether the user agrees to authorize or disagrees with authorizing. This example is a scenario where the user agrees to authorize. The application result needs to be uploaded to the system for recording.

[0133] It requests permission from the terminal device's operating system to read database files in the local storage space and collects the request result from the system.

[0134] S61: If the application result is "authorization granted", then scan the database file and identify candidate game information through the preset game type classification table.

[0135] Candidate game information refers to game information to be retrieved in local storage.

[0136] The game type classification table is a reference table that is pre-defined by those skilled in the art to classify various types of game data into corresponding game types.

[0137] If the application result is "authorization approved", it means that the user agrees to the authorization. The database file needs to be scanned by the file scanning engine, and the scanned content will be identified through the game type classification table to obtain candidate game information.

[0138] S62: Retrieve game log information based on candidate game information.

[0139] Game log information refers to the log information stored locally in the game.

[0140] The game log information can be retrieved based on the candidate game information using the log retrieval module.

[0141] S63: Extract game log information to obtain locally stored game information.

[0142] Locally stored game information refers to game-related information stored locally on the player's client.

[0143] Key fields are extracted from game log information, and retrieval is performed based on these key fields to obtain locally stored game information, thus achieving the effect of legally obtaining similar game behavior data with user authorization.

[0144] It also includes methods for processing profiles of players who have been inactive for a long time: S70: Traverse the profile records corresponding to each biological ID in the biological behavior profile database and extract the last active timestamp of each profile record.

[0145] The last active timestamp refers to the system time when the player's behavioral profile was last retrieved.

[0146] The database traversal engine iterates through the profile records corresponding to each biological ID in the biological behavior profile database and extracts the last active timestamp of each profile record.

[0147] S71: Collect the current system time and calculate the difference between the current system time and the last active timestamp to obtain the inactivity duration of the profile.

[0148] Inactive time refers to the time during which a player's behavioral profile is not invoked.

[0149] The system clock module collects the current system time, and the difference between the current system time and the last active timestamp is calculated to obtain the inactive duration.

[0150] S72: When the inactive time exceeds the preset quit time threshold, mark the biometric ID as a quitting player and mark the corresponding player behavior profile as a profile to be compressed.

[0151] The profile to be compressed refers to the player behavior profile that needs to be compressed.

[0152] The game quitting time threshold refers to a time length value preset by those skilled in the art to determine whether a player has not logged into the game for a long time. When the last active time of a certain biometric ID exceeds the current system time, the player is marked as a quitting player.

[0153] The inactivity duration is compared with the quitting time threshold. When the threshold is exceeded, the biometric ID is marked as a quitting player through the tagging module, and its profile is marked as a profile to be compressed.

[0154] S73: Perform compression processing on the compressed image and re-store the compressed image into the biological behavior image library.

[0155] The compressed profile is compressed while retaining the basic preference labels and core weight vectors, and deleting the behavioral sequence data. The compressed profile is then stored back into the biological behavioral profile library, which frees up storage resources.

[0156] Basic preference tags refer to the tags for preference content in a player's behavioral profile.

[0157] The core weight vector refers to the data content with the highest weight value in the player behavior profile.

[0158] S74: When a player who has been marked as a quitter is detected to log in again through biometric identification, the player's profile is decompressed and the real-time updates of the profile are restored.

[0159] When a player's bio-ID, which has been marked as a quitter, is searched again in the bio-behavioral profile database, it indicates that the profile corresponding to that player needs to be decompressed and the real-time updates of that profile need to be restored.

[0160] It also includes a smart resource loading method for returning players: S80: When a player who has been marked as a quitting player is detected to log in again through biometric identification, that player is marked as a returning player.

[0161] Returning players are those who were previously marked as quitting the game but then log in again using biometric identification.

[0162] When a player who has left the game searches for their biometric ID again, it indicates that the player is a returning player and should be marked as such.

[0163] S81: Collect the biometric data of returning players to obtain their IDs.

[0164] Returning player ID refers to the avatar ID that is marked as a returning player.

[0165] Similarly, the biometric data of returning players are collected through step S10 above, and the returning player ID is obtained through one-way hash operation.

[0166] S82: Based on the returning player ID, the biological behavior profile database is traversed and retrieved to obtain the returning player behavior profile.

[0167] Returning player behavior profile refers to the behavior profile of players who have been marked as returning players.

[0168] The behavioral profiles of returning players are obtained by traversing and searching the biometric profile database based on their IDs.

[0169] S83: Collect promotional information from the game's official website within a preset time period and keywords from discussions about the game on mainstream social networking sites, and compare them with returning player behavior profiles to obtain player return keywords.

[0170] Promotional information refers to the game information released by this game.

[0171] Keywords refer to popular terms used in discussions about the game on mainstream social networking sites.

[0172] The preset time period refers to a time period set in advance by those skilled in the art, within one month of the current system time.

[0173] Player return keywords refer to the key factors that may lead to a player's return.

[0174] By using web crawlers to collect promotional information from the game's official website within a preset time period and keywords related to discussions about the game on mainstream social networking sites, and matching them with basic preference tags from returning player behavior profiles, the game's returning player keywords can be retrieved.

[0175] S84: Extract the resource package information to be loaded based on player return keywords to obtain player return resource information, and execute the priority loading method for game resources according to the player return resource information.

[0176] Player return resource information refers to the portion of the resource package information that matches the player return keyword.

[0177] By comparing and retrieving information on resource packages that need to be loaded based on player return keywords, the system obtains player return resource information and prioritizes loading game resources, thereby achieving the effect of accurately preloading content that returning players are interested in.

[0178] It also includes an iterative update method for the initial behavioral profile: S90: Collects the number of game sessions for new players.

[0179] The number of new player game sessions refers to the number of times a new player plays the game.

[0180] The number of game sessions for new players is collected through the device terminal, and the number of game sessions is incremented by one after each game ends.

[0181] S91: When the number of game sessions for a new player exceeds the preset threshold for the number of game sessions, collect the personal behavior sequence data generated by the new player in the game session.

[0182] The game session count threshold refers to a game session count value preset by those skilled in the art, used to trigger the first iterative update of the initial behavioral profile of a new player.

[0183] Once a new player completes more than the threshold number of game sessions, it indicates that the new player has accumulated enough data, and the system needs to initiate the profile calibration process to collect personal behavioral sequence data.

[0184] Personal behavior sequence data refers to a collection of a series of operational behavior events recorded in chronological order by a new player in the game. Each event includes the behavior type, the timestamp of occurrence, and the current state or interface identifier.

[0185] By processing individual behavioral sequence data using conditional probability models, a unique behavioral profile is generated, reflecting the player's true preferences in the game.

[0186] S92: Input individual behavioral sequence data into a conditional probability model to generate a unique behavioral profile for new players.

[0187] The exclusive behavioral profile for new players refers to the profile generated for new players during the cold start phase. After the new player completes a preset number of game sessions and accumulates sufficient personal data, the system analyzes their personal behavioral sequence data through a conditional probability model to generate a profile that reflects the player's real behavioral habits and content preferences in the game.

[0188] S93: Calculate the difference between the customized behavioral profile and the initial behavioral profile.

[0189] The difference refers to the difference between the corresponding data of the exclusive behavioral profile and the initial behavioral profile.

[0190] The cosine similarity algorithm is used to calculate the degree of deviation between the new player's personalized behavioral profile and the initial behavioral profile.

[0191] S94: When the difference exceeds the preset profile difference threshold, replace the initial behavioral profile with a unique behavioral profile, and then re-bind the unique behavioral profile with the new player's biometric ID before uploading it to the biometric behavioral profile library on the cloud server.

[0192] The profile difference threshold refers to a standard value preset by those skilled in the art to determine whether the deviation between the initial behavioral profile and the personalized behavioral profile is significant. When the difference calculated by the cosine similarity algorithm is higher than this threshold, it indicates that the initial behavioral profile can no longer accurately reflect the player's true preferences. The initial behavioral profile needs to be replaced with a personalized behavioral profile, and the personalized behavioral profile should be re-bound to the new player's biometric ID before being uploaded to the biometric behavioral profile library on the cloud server.

[0193] Based on the same inventive concept, embodiments of the present invention provide a cross-account game resource intelligent loading system based on biometric recognition, comprising: The data collection module is used to collect player biometric data, player behavior data, spatial heat vectors, real-time player biometric data, various content modules, dwell time information, historical behavior data, real-time behavior data, locally stored game information, resource package information to be loaded, application results, current system time, biometric data of returning players, promotional information, keywords, number of game sessions of new players, and personal behavior sequence data. The memory is used to store the program that implements a cross-account game resource intelligent loading method based on biometric recognition; The processor is used to load and execute programs stored in memory.

[0194] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. 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. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0195] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent loading of cross-account game resources based on biometric recognition, characterized in that, include: Collect player biometric data, player behavior data, and spatial heat vectors; Player biometric IDs are obtained by identifying player biometric data; Player behavior data and spatial heat vectors are input into a preset conditional probability model to obtain a player behavior profile; Upload player biometric IDs and player behavioral profiles to the cloud server and bind them to obtain a biometric behavioral profile database; Collect real-time player biometric data during account login; When real-time player biometric data is inconsistent with player biometric data, extract the real-time player biometric ID based on the real-time player biometric data; Real-time player biometrics are obtained by retrieving and matching real-time player biometrics with a biometrics behavior profile database. Based on real-time player behavior profiles, the game resources are loaded using a pre-defined priority loading method to obtain a priority loading sequence and an idle asynchronous loading queue, and then the game resources are loaded.

2. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 1, characterized in that, It also includes methods for acquiring spatial thermal vectors: Collect data from each content module; Each content module is equally divided into multiple grid units to obtain the grid units for each module; Collect information on the duration of player stay in each module's grid cell; The thermal values ​​of grid cells are extracted based on the dwell time information; The spatial thermal vector is obtained by performing ensemble analysis on the thermal values ​​of the grid cells.

3. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 1, characterized in that, It also includes methods for establishing conditional probability models: Collect historical behavioral data from player accounts; The state transition frequency requirement is obtained by calculating the state transition frequency based on historical behavior data. The conditional probability model is trained by the state transition frequency requirement; Collect players' real-time behavior data; Compare players' real-time behavior data with their historical behavior data to obtain player behavior difference values ​​and accumulate them; When the accumulated player behavior difference value exceeds the preset behavior difference threshold, the state transition frequency requirement is collected and updated again, and the parameters of the conditional probability model are updated based on the updated state transition frequency requirement.

4. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 1, characterized in that, It also includes a cold start loading method for new players with no historical behavior data: When no real-time player behavior profile corresponding to the real-time player's biological ID is found in the biological behavior profile database, the player is determined to be a new player, and the game information stored locally on the terminal device used by the new player is collected. Extract local behavior data from game information stored locally; Extract behavioral preference features from local behavioral data; An initial behavioral profile of the new player is generated based on behavioral preference characteristics. This initial behavioral profile is then bound to the real-time player's biometric ID and uploaded to the biometric profile library on the cloud server. Game resources are then loaded with priority based on the initial behavioral profile.

5. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 4, characterized in that, The priority loading method includes: Collect information about the resource packages that need to be loaded; Identify player behavior preferences based on real-time player behavior profiles; By extracting the information of the resource packs that need to be loaded from the player's behavioral preference information, candidate resource packs and non-candidate resource packs are obtained; Extract the frequency and timing of candidate resource packs in player behavior preference information; The priority weight value is obtained by weighting the candidate resource packs based on their frequency and time of appearance in the player's behavioral preference information. All candidate resource packages are sorted from highest to lowest priority weight to obtain a priority loading queue, and priority loading is performed according to the priority loading queue. Non-candidate resource packages are uniformly arranged into an idle asynchronous loading queue.

6. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 4, characterized in that, The collection of game information locally stored on the terminal device used by new players specifically includes: Request the terminal device's operating system to read the database file from the local storage space and collect the request result; If the application result is "authorization granted", the database file is scanned and candidate game information is obtained by identifying the game type classification table. Retrieve game log information based on candidate game information; Game log information is extracted to obtain locally stored game information.

7. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 1, characterized in that, It also includes methods for processing profiles of players who have been inactive for a long time: Iterate through the bio profile records corresponding to each bio ID in the bio behavior profile database and extract the last active timestamp of each bio profile record; Collect the current system time and calculate the difference between the current system time and the last active timestamp to obtain the inactivity duration of the profile; When the inactive time exceeds the preset quit time threshold, the biometric ID is marked as a quitting player, and the corresponding player behavior profile is marked as a profile to be compressed. The compressed image is compressed and then stored back into the biological behavior image database. When a player who has been marked as a quitter is detected to log in again via biometric identification, the player's profile is decompressed and the real-time updates of that profile are restored.

8. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 7, characterized in that, It also includes a smart resource loading method for returning players: When a player who has been marked as a quitter is detected to log in again via biometric identification, that player is marked as a returning player. Collect the biometric data of returning players to obtain their IDs; The behavioral profiles of returning players are obtained by traversing and searching the biometric profile database based on the returning player IDs. Collect promotional information from the game's official website within a preset time period and keywords related to discussions about the game on mainstream social networking sites, and compare them with returning player behavior profiles to obtain player return keywords; Based on player return keywords, the resource package information to be loaded is extracted to obtain player return resource information, and the game resources are loaded according to priority based on the player return resource information.

9. The method for intelligent loading of cross-account game resources based on biometric recognition according to claim 4, characterized in that, It also includes an iterative update method for the initial behavioral profile: Collect the number of game sessions for new players; When the number of game sessions for a new player exceeds the preset threshold, collect the personal behavior sequence data generated by the new player during the game session; Input individual behavioral sequence data into a conditional probability model to generate a unique behavioral profile for new players; Calculate the difference between the customized behavioral profile and the initial behavioral profile; When the difference exceeds the preset profile difference threshold, the initial behavioral profile is replaced with a unique behavioral profile, and the unique behavioral profile is re-bound to the new player's biometric ID before being uploaded to the biometric behavioral profile library on the cloud server.

10. A cross-account game resource intelligent loading system based on biometric recognition, characterized in that, include: The data acquisition module is used for player biometric data, player behavior data, spatial heat vectors, and real-time player biometric data. A memory for storing a program that implements a cross-account game resource intelligent loading method based on biometric recognition as described in any one of claims 1 to 9; The processor is used to load and execute programs stored in memory.

Images