A self-learning virtual display method and system based on external stimulus input

Abstract

The present application relates to the technical field of virtual games, and discloses a self-learning virtual display method and system based on external stimulus input, including: real-time monitoring to obtain the game process status and the external stimulus input data after normalization and quantization processing; generating corresponding original image data according to the game process status and the external stimulus input data; configuring a selection box to select the original image data to obtain sampled image data; inputting the sampled image data into a trained mode recommendation model to obtain the classification result output by the mode recommendation model; configuring the game virtual mode according to the classification result; being able to provide the original mode recommendation model with a large number of training samples with flexible and changeable specifications during the training phase, and combining with a large number of sampled image data of players obtained from the cloud, so as to steadily improve the classification accuracy, take into account the personalized needs of players during classification, and comprehensively improve the game experience of players, so that players can get a better game experience in the game.

Description

Technical Field

[0001] The present application relates to the technical field of virtual games, and in particular to a self-learning virtual display method and system based on external stimulus input. Background Art

[0002] Metaverse games are a type of virtual world created by scientific and technological means with blockchain as the underlying technology support. They have mapping and interactive functions with the real world and can provide players with immersive experience, open world exploration, social interaction, economic system, user-generated content, and cross-platform compatibility. The immersive experience provided in metaverse games can be achieved through advanced graphics technology and interactive devices, such as VR and AR technology, so that players can have an immersive gaming experience. This sense of immersion makes the game world more real, and the players' sense of participation and involvement are greatly improved.

[0003] In metaverse games, there is usually a vast virtual space where players can move freely, without being restricted by traditional game levels or tasks, which can provide players with more freedom and exploration space, making the game experience richer and more diverse. The virtual modes of metaverse games generally include NPC interaction and plot branching strategies, virtual environment simulation strategies, and enemy behavior patterns and strategies. In terms of NPC interaction and plot branching strategies, metaverse games can provide more humanized NPC characters and richer and more diverse plot branches due to their freedom. NPC characters can have unique personalities, backgrounds and stories, and can show different emotional reactions according to the player's behavior and words. The multiple plot branches can also fully meet the player's own emotional value, making players more willing to experience the game in depth, so as to continuously experience richer and more diverse game content; in terms of virtual environment simulation strategies, the game environment can change dynamically, such as weather, seasons, terrain, etc., so as to further increase the immersion and realism of the game; in terms of enemy behavior patterns and strategies, enemy units of different difficulty levels can be provided for players to attack, meeting the gameplay and operability requirements of different player groups.

[0004] However, the configuration adjustment of the game virtual mode of the above-mentioned metaverse game requires players with rich gaming experience. Players who are just starting to get in touch with metaverse game types need to spend a certain amount of time learning and adapting to such games. In addition, some players do not have sufficient gaming time and just want to experience the fun of the game rather than challenges, real simulations, etc. When using wearable devices to play games, the environmental simulation in some situations may be too realistic, which will cause discomfort to some players and increase the burden on the device operation, affecting the smoothness of the game operation. There are many plot branches and NPC interaction strategy contents. First, it is not suitable for players to make manual decisions on their own. Second, if the players choose to define them by themselves, there will be more subjective nature, and the emotional value brought by randomness will be lost, which is not conducive to players trying the game for a long time and many times. Moreover, in a combat state, the enemy's attack behavior or strategy may not be able to automatically and dynamically take care of all players, which also affects the player's gaming experience. Therefore, how to dynamically adjust the personalized game virtual mode for players while not increasing the burden on the device operation too much has become a technical problem that needs to be solved at present. Summary of the invention

[0005] In order to provide various types of players with a more personalized, more immersive and better overall gaming experience, the present application provides a self-learning virtual display method and system based on external stimulus input.

[0006] In a first aspect, the present application provides a self-learning virtual display method and system based on external stimulus input, which adopts the following technical solutions:

[0007] A self-learning virtual display method based on external stimulus input, comprising:

[0008] S1, real-time monitoring to obtain the game process status and the external stimulus input data after normalization and quantization processing;

[0009] S2, generating corresponding original image data according to the game process state and the external stimulus input data;

[0010] S3, configuring a selection box to select the original image data to obtain sampled image data;

[0011] S4, inputting the sampled image data into the trained mode recommendation model to obtain the classification result output by the mode recommendation model;

[0012] S5, configuring the game virtual mode according to the classification result;

[0013] Among them, the game process status includes plot state, free state and combat state; the external stimulus input data includes user physiological data, environmental parameter data and user operation data; the game virtual mode includes NPC interaction and plot branching strategy, virtual environment simulation strategy and enemy behavior mode and strategy; the original image data includes n image partitions with sorting rules, and the image partitions are correspondingly associated with the data categories of the external stimulus input data, and the sorting rules represent the order of weights of the data categories of the external stimulus input that affect the classification results.

[0014] In this way, the present invention converts the external stimulus input data that can reflect the user's physiological data, environmental parameter data and user operation data together with the game process status into original image data, and unifies different types of data in the form of image data, thereby providing a basis for rapid classification for the mode recommendation model trained based on the neural network model, and then selects the sampled image data suitable for the mode recommendation model in the form of a selection box, which can provide the original mode recommendation model with a large number of training samples with flexible and changeable specifications during the training stage, and can be combined with a large number of sampled image data of players obtained from the cloud, so as to achieve a steady improvement in the classification accuracy, and at the same time, it can also take into account the personalized needs of players in the classification, and comprehensively improve the game experience of players, so that players can get a better game experience in the game.

[0015] Optionally, step S2 includes:

[0016] S21, configuring a blank picture of a specified size;

[0017] S22, dividing the blank picture into a process identification area and a data identification area;

[0018] S23, dividing the data identification area into n first-level image partitions;

[0019] S24, dividing each of the first-level picture partitions into corresponding second-level picture partitions;

[0020] S25, filling the external stimulus input data into the secondary image partition according to the subdivided categories, and when the filling progress reaches a preset progress threshold, proceeding to step S26;

[0021] S26, sorting the n first-level image partitions according to the sorting rule;

[0022] Among them, the filling pixel hues of different first-level image partitions are different, the filling pixel hues of the second-level image partitions of the same first-level image partition are the same but the color depths are different, and the numerical size of the subdivided category data is represented by the brightness of the filling pixels.

[0023] Optionally, the sorting rules include:

[0024] Calculate the subdivided data fluctuation variance V corresponding to the subdivided category data of each data category of the external stimulus input data ij , and according to the segmented data fluctuation variance V ij Calculate the average fluctuation variance V corresponding to each data category of the external stimulus input data i ; Wherein, i is the number corresponding to the first-level image partition, and j is the number of the second-level image partition corresponding to the subdivided category data;

[0025] According to the average volatility variance V i Sort the first-level image partitions according to the variance V of the segmented data fluctuations ij The secondary picture partitions are sorted within the primary picture partitions.

[0026] Optionally, the segmented data fluctuation variance V ij and the average volatility variance V i The methods for obtaining include:

[0027]

[0028] Among them, x ijk is the pixel brightness value of the k-th pixel in the j-th secondary image partition of the ith primary image partition, is the average brightness of pixels in the jth secondary image partition of the ith primary image partition, m ij is the total number of padded pixels in the jth secondary image partition of the ith primary image partition, p i is the total number of secondary image partitions of the i-th primary image partition.

[0029] Optionally, the method for configuring the selection box includes:

[0030] Calculate and obtain the equipment load assessment coefficient U;

[0031] According to the equipment load evaluation coefficient U and the comprehensive fluctuation variance V, the selection area S, the selection content overlap C and the selection frequency f of the selection box are configured;

[0032] Among them, the method for obtaining the comprehensive volatility variance V includes:

[0033]

[0034] The equipment load evaluation coefficient U is negatively correlated with the cut-off area S, the cut-off content overlap C, and the cut-off frequency f;

[0035] The comprehensive fluctuation variance V is positively correlated with the cut-off area S, the overlap degree C of the cut-off content, and the cut-off frequency f.

[0036] By adopting the above technical solution,

[0037] Optionally, the method for obtaining the equipment load evaluation coefficient U includes:

[0038] Obtain device-related operating parameters, including CPU occupancy rate R1, GPU occupancy rate R2, memory usage rate R3, game refresh rate pressure value R4, and key hardware position temperature pressure value R5;

[0039]

[0040]

[0041] Among them, X fMAX is the maximum supported refresh rate, X f(t) is the average refresh rate of the current t-th unit period, X fMAX and X fMIN They are the historical maximum refresh rate and historical minimum refresh rate in the same game process state, X W(t) and X W(t-1) are the average temperature of the current t-th unit period and the average temperature of the t-1-th unit period, respectively. WMAX and X WMIN are the historical maximum key position temperature and the historical minimum key position temperature in the same game process state, Kf is the preset compensation frame number, R 4SD and R 5SD All are preset pressure constants.

[0042] Optionally, the method for configuring the selection area S, the selection content overlap C, and the selection frequency f of the selection box includes:

[0043] The configuration method of the cut-off area S includes:

[0044]

[0045] The configuration method of the overlap degree C of the selected content includes:

[0046]

[0047] The configuration method of the cut-off frequency f includes:

[0048]

[0049] Among them, S Max and S Minare the maximum and minimum cross-sectional areas, respectively, and C Max and C Min are the maximum overlap of selected content and the minimum overlap of selected content, respectively, and f Max and f Min are the maximum and minimum cut-off frequencies, μ1, μ2, and μ3 are preset adjustment coefficients, S∈[S Min ,S Max ], C∈[C Min ,C Max ], f∈[f Min ,f Max ],V SD , U SD They are respectively a preset variance threshold and a preset load critical threshold.

[0050] Optionally, the sampled image data includes at least one sampled data unit, and each sampled data unit corresponds to primary image data selected by the selection box.

[0051] Optionally, when the game process state is switched, a weak reminder setting inquiry box of the game virtual mode is sent to the user;

[0052] Adjusting the virtual mode of the game according to the user's intention and setting result;

[0053] The mode recommendation model is supplementarily trained according to the intention and the setting result, and the mode recommendation model after the supplementary training is uniquely associated with the user.

[0054] In a second aspect, the present application provides a self-learning virtual display system based on external stimulus input, which adopts the following technical solution:

[0055] A self-learning virtual display system based on external stimulus input, wherein the self-learning virtual display system runs a program of any one of the self-learning virtual display methods described above.

[0056] The self-learning virtual display system includes:

[0057] The data acquisition module is used to monitor and obtain the game process status and the external stimulus input data after normalization and quantification in real time;

[0058] A data generation module, used to generate corresponding original image data according to the game process state and the external stimulus input data;

[0059] A data selection processing module, used for configuring a selection box, and obtaining sampled image data by selecting the original image data through the selection box;

[0060] A data sending module, used for inputting the sampled image data into a trained mode recommendation model to obtain a classification result output by the mode recommendation model; the mode recommendation model is a classification neural network model after training;

[0061] A terminal auxiliary configuration module, used to configure the game virtual mode according to the classification result;

[0062] Among them, the external stimulus input data includes user physiological data, environmental parameter data and user operation data; the game process status includes plot status, free status and combat status; the game virtual mode includes NPC interaction and plot branching strategy, virtual environment simulation strategy and enemy behavior mode and strategy; the original image data includes n image partitions with sorting rules, and the image partitions are correspondingly associated with the data categories of the external stimulus input data, and the sorting rules represent the order of weights of the data categories of the external stimulus input that affect the classification results.

[0063] In summary, the present application includes at least one of the following beneficial technical effects:

[0064] 1. The present invention converts the external stimulus input data that can reflect the user's physiological data, environmental parameter data and user operation data together with the game process status into original image data, and unifies different types of data in the form of image data, thereby providing a basis for rapid classification for the mode recommendation model trained based on the neural network model, and then selects the sampled image data suitable for the mode recommendation model in the form of a selection box, which can provide the original mode recommendation model with a large number of training samples with flexible and changeable specifications during the training stage, and can be combined with a large number of sampled image data of players obtained from the cloud, so as to achieve a steady improvement in the classification accuracy, and at the same time, it can also take into account the personalized needs of players in the classification, and comprehensively improve the game experience of players, so that players can get a better game experience in the game.

[0065] 2. The present invention utilizes sorting rules to enable the trained pattern recommendation model to "notice" the change and reflect it in the output of the classification result when a certain sub-category of the input data is greatly mutated due to external stimulation of the user. Such a design can provide more suitable labels for annotation during the training phase, providing a basis for the annotation and classification of complex scenes. It can also effectively multiply the training sample data and improve the classification efficiency and quality across the board.

[0066] 3. The present invention incorporates the device load assessment coefficient U and the comprehensive fluctuation variance V into the configuration of the selection box at the same time, which can take into account the device computing power load management while ensuring the doubling of training samples, reduce the operating burden of the device caused by the operation of the mode recommendation model, reduce the cost of building player equipment, and ensure the normal basic game experience of the player. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is the overall flow chart of the self-learning virtual display method based on external stimulus input in the present invention. DETAILED DESCRIPTION

[0068] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings.

[0069] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0070] The present application embodiment discloses a self-learning virtual display method based on external stimulus input, referring to Figure 1 ,include:

[0071] S1, real-time monitoring to obtain the game process status and the external stimulus input data after normalization and quantization processing;

[0072] S2, generating corresponding original image data according to the game process status and external stimulus input data;

[0073] S3, configure a selection box to select the original image data to obtain the sampled image data;

[0074] S4, inputting the sampled image data into the trained mode recommendation model to obtain the classification result output by the mode recommendation model;

[0075] S5. configuring the game virtual mode according to the classification result;

[0076] Among them, the game process status includes plot status, free status and combat status; the external stimulus input data includes user physiological data, environmental parameter data and user operation data; the game virtual mode includes NPC interaction and plot branching strategies, virtual environment simulation strategies and enemy behavior patterns and strategies; the original image data includes n image partitions with sorting rules, and the image partitions are correspondingly associated with the data categories of the external stimulus input data. The sorting rules represent the order of the weights of the data categories of the external stimulus input that affect the classification results.

[0077] In an exemplary embodiment of the present invention, the user physiological data may include heart rate data, blood pressure data, respiratory rate, blood oxygen data, plot attention data, etc.; the environmental parameter data may include local environmental data, user environmental data and device environmental data; local environmental data may include local weather conditions and day and night conditions, user environmental data may include the user's environment conditions, such as temperature, humidity and noise level, etc., and device environmental data may include the temperature of key positions of the device and the frequency of use of various devices (such as VR / AR head display devices, tactile devices, PC peripheral devices, etc.);

[0078] User operation data may include the frequency of viewing guides, the frequency of effective use of game equipment and items, player combat evaluation scores, player chat frequency, NPC interaction frequency, transaction reset frequency, player rating data inside and outside the game, player feedback frequency, etc.

[0079] Among them, in the process of normalization and quantization, taking heartbeat data as an example, according to the player's historical heartbeat data, the maximum heartbeat data and minimum heartbeat data within the recent period, such as 5 minutes, can be obtained, and then normalized according to the proportion of the current heartbeat data exceeding the minimum heartbeat data. If the current heartbeat data is lower than the minimum heartbeat data or higher than the maximum heartbeat data, special numerical definitions can be directly made to distinguish them; in addition, plot attention data and other such data cannot be directly collected and obtained. The user's attention to the plot content or NPC dialogue tab can be determined based on eye movement detection and device display content, and then the corresponding attention preset value is adjusted. If the user does not pay attention to the relevant designated position during the plot process or the NPC dialogue process, the attention preset value is lowered, otherwise the attention preset value is increased to achieve unified quantization. Here, the applicant points out that there is no need to specify the normalization and quantization method for different data. It is only necessary to simulate the change trend and amplitude of the original data through normalization and quantization, and no special requirements are made on the simulation accuracy.

[0080] Based on the above technical content, this illustrative implementation converts the external stimulus input data that can reflect the user's physiological data, environmental parameter data and user operation data together with the game process status into original image data, and unifies different types of data in the form of image data, thereby providing a basis for rapid classification for the mode recommendation model trained based on the neural network model, and then selecting the sampled image data suitable for the mode recommendation model in the form of a selection box, which can provide the original mode recommendation model with a large number of training samples with flexible and changeable specifications during the training stage, and can be combined with a large number of sampled image data of players obtained from the cloud, to achieve a steady improvement in classification accuracy, while also being able to take into account the personalized needs of players during classification, and comprehensively improve the player's gaming experience, so that players can get a better gaming experience in the game.

[0081] The key point of this exemplary embodiment is that it adopts a technical route of partition sorting and box selection of sampled image data with category partitions. In the training process, since the acquisition method of training samples and sampled image data is the same, only the step of labeling is added. The multi-type data filling in the category partition after normalization and quantization can ensure the diversification of model classification results (that is, labels). The partition sorting adopts a preset sorting rule, which can achieve that even if the data in the partition is very similar (for example, the similarity is more than 99.9%), the sampled image data may be given different labels during training due to the different partition arrangement order, so that the pattern recommendation model outputs a large difference in classification results; and the design of the box selection data can utilize the changes in the box selection area and the box selection position. In the stage of generating training samples, theoretically exponentially similar training samples can be generated for the same original image data. In this way, the accuracy of the model's later recognition can be improved even when limited original image data is obtained, interference factors can be reduced, and the reliability and robustness of the model recognition can be fully improved.

[0082] Optionally, step S2 includes:

[0083] S21, configuring a blank picture of a specified size;

[0084] S22, dividing the blank image into a process identification area and a data identification area;

[0085] S23, dividing the data identification area into n first-level image partitions, where n can be 3;

[0086] S24, dividing each first-level picture partition into corresponding second-level picture partitions;

[0087] S25, filling the external stimulus input data into the secondary image partition according to the subdivided categories, and when the filling progress reaches a preset progress threshold, proceeding to step S26;

[0088] S26, sorting the n first-level image partitions according to the sorting rule;

[0089] Among them, the filling pixel hues of different first-level image partitions are different, the filling pixel hues of the second-level image partitions of the same first-level image partition are the same but the color depths are different, and the numerical size of the subdivided category data is represented by the brightness of the filling pixels.

[0090] In this exemplary embodiment of the present invention, the mode recommendation model is preferably a convolutional neural network (CNN), the input layer is set to RGB three channels for receiving color images, and a normalization layer is provided.

[0091] Similarly, taking the heartbeat data as an example, the heartbeat data can be associated with the first secondary image partition corresponding to the first primary image partition. Under the premise that the user's physiological data is selected as a red hue, the secondary image partition where the heartbeat data is located can be filled with 50% light red when filling pixels, and the brightness is set accordingly according to the size of the normalized and quantized heartbeat data; in this way, the first heartbeat data after normalization and quantization can be filled with the first pixel in the upper left corner of the first secondary image partition, so that the pixel filling of the second heartbeat data can be completed to the right. After the first row of pixels is filled, it can be transferred to the first pixel of the next row until the filling progress reaches 30% (that is, the total number of filled pixels reaches 30% of the secondary image partition), and the first-level image partition can be sorted according to the sorting rules. In addition, when the game process state changes, the process identification area changes synchronously, and the pixels in the secondary image partition that have not been filled before the process identification area changes can be uniformly filled with white.

[0092] Such a design can densely integrate a variety of different types of external stimulus input data into image data that can be recognized by the neural network model, use hue and color depth to distinguish categories, and use brightness differences to achieve efficient expression of data in time series changes. In this way, when there is a large amount of changing data of different categories, the characteristics of the neural network model can be fully utilized, and when there are the same number of label types, the corresponding classification results can be output efficiently and accurately.

[0093] Optionally, in this exemplary embodiment, the sorting rules include:

[0094] Calculate the variance V of the subdivided data corresponding to the subdivided data of each external stimulus input data category ij , and according to the segmented data fluctuation variance V ij Calculate the average fluctuation variance V corresponding to each data category of the external stimulus input data i; Where i is the number of the corresponding first-level image partition, j is the number of the second-level image partition corresponding to the subdivided category data; In this embodiment, the heartbeat data can be taken as an example, and the variance of the subdivided data fluctuation is V 11 , then the average fluctuation variance corresponding to the user's physiological data is V1, and the average fluctuation variance corresponding to the environmental parameter data can be V2.

[0095] Segment data volatility V ij and the average volatility variance V i The methods for obtaining include:

[0096]

[0097] Among them, x ijk is the pixel brightness value of the k-th pixel in the j-th secondary image partition of the ith primary image partition, is the average brightness of pixels in the jth secondary image partition of the ith primary image partition, m ij is the total number of padded pixels in the jth secondary image partition of the ith primary image partition, p i is the total number of secondary image partitions of the i-th primary image partition.

[0098] After the calculation is completed, according to the average volatility variance V i Sort the first-level image partitions according to the variance V of the segmented data ij Sort the second-level image partitions within the first-level image partition. If V2>V1, the sorting can be arranged from left to right in the order of process identification area, second-level image partition, and first-level image partition; if V 12 >V 11 , when arranging the secondary picture partitions in the first primary picture partition, from left to right they are the second secondary picture partition and the first secondary picture partition.

[0099] With this setting, when a large mutation occurs in the user's heartbeat data, the trained pattern recommendation model can "notice" the change and reflect it in the output of the classification result. This design can provide labels with a higher degree of fit during the training phase, providing a basis for the annotation and classification of complex scenes. It can also effectively multiply the training sample data and improve the classification efficiency and quality across the board.

[0100] Optionally, the configuration method of the selection box includes:

[0101] Calculate and obtain the equipment load assessment coefficient U;

[0102] According to the equipment load evaluation coefficient U and the comprehensive fluctuation variance V, the selection area S, the selection content overlap C and the selection frequency f of the selection box are configured;

[0103] Among them, the method for obtaining the comprehensive volatility variance V includes:

[0104]

[0105] The equipment load assessment coefficient U is negatively correlated with the cut-off area S, the cut-off content overlap C, and the cut-off frequency f;

[0106] The comprehensive volatility variance V is positively correlated with the cut-off area S, the overlap of the cut-off content C, and the cut-off frequency f.

[0107] By adopting the above technical solution, the device load assessment coefficient U and the comprehensive fluctuation variance V are simultaneously incorporated into the configuration of the selection box, which can ensure the doubling of training samples while taking into account the device computing load management, reduce the operating burden of the device caused by the operation of the mode recommendation model, reduce the cost of building player equipment, and ensure the normal basic game experience of the player; in this embodiment, the sampled image data includes at least one sampling data unit, and each sampling data unit corresponds to the image data intercepted by the selection box.

[0108] With such settings, if the equipment load assessment coefficient U is higher, it means that the equipment operating pressure is greater. At this time, reducing the interception area S can reduce the amount of data received by the mode recommendation model, and reducing the interception content overlap C and the interception frequency f can reduce the degree and number of data repetitions of adjacent sampling data units, making full use of limited computing power for timely classification; if the comprehensive fluctuation variance V is larger, it means that the fluctuation of the input data due to external stimulation is larger, and sampling data units with larger data content are required to ensure the accuracy of classification, and the continuity of adjacent sampling data units can be improved by improving the interception content overlap C, which helps the sampling data units to give accurate classification results when facing large data fluctuations in the input data due to external stimulation.

[0109] Specifically in this embodiment of the present invention, the method for obtaining the equipment load evaluation coefficient U includes:

[0110] Get device-related operating parameters, including CPU occupancy rate R1, GPU occupancy rate R2, memory usage rate R3, game refresh rate pressure value R4, and key hardware location temperature pressure value R5;

[0111]

[0112] Among them, X fMAX The maximum supported refresh rate is usually determined by the display device or graphics card. f(t) is the average refresh rate of the current t-th unit period, X fMAX and X fMIN They are the historical maximum refresh rate and historical minimum refresh rate in the same game process state, XW(t) and X W(t-1) are the average temperature of the current t-th unit period and the average temperature of the t-1-th unit period, respectively. WMAX and X WMIN are the historical maximum key position temperature and the historical minimum key position temperature in the same game process state, Kf is the preset compensation frame number, to avoid (X f(t) -X fMIN )+Kf is 0, R 4SD and R 5SD All are preset pressure constants. Under the condition of adaptive refresh rate setting, X f(t) The larger it is, the lower the pressure on the current image quality operation; Its exponential property can be used to W(t) Close to or greater than X WMAX The R5 value can be increased rapidly at the same time, thereby reducing the operating pressure of the equipment more accurately and predictably, and better protecting expensive gaming equipment.

[0113] Optionally, the method for configuring the selection area S, the selection content overlap C, and the selection frequency f of the selection box includes:

[0114] The configuration method of the cut-off area S includes:

[0115]

[0116] The configuration method of the overlap degree C of the selected content includes:

[0117]

[0118] The configuration method of the cut-off frequency f includes:

[0119]

[0120] Among them, S Max and S Min are the maximum and minimum cross-sectional areas, respectively, and C Max and C Min are the maximum overlap of selected content and the minimum overlap of selected content, respectively, and f Max and f Min are the maximum and minimum cut-off frequencies, μ1, μ2, and μ3 are preset adjustment coefficients, S∈[S Min ,S Max ], C∈[C Min ,C Max ], f∈[f Min ,f Max ],V SD , U SD They are respectively a preset variance threshold and a preset load critical threshold.

[0121] In addition, in another embodiment of the present invention, when the game process state is switched, a weak reminder setting inquiry box of the game virtual mode may be sent to the user, such as a short inquiry of 3 seconds popping up in a bubble in the corner of the screen;

[0122] According to the user's willingness and setting results of adjusting the game virtual mode;

[0123] The mode recommendation model is supplemented with training according to the intention and setting results, and the supplemented mode recommendation model is uniquely associated with the user. This can fully realize personalized game mode customization switching service and provide players with a better overall gaming experience.

[0124] In a second aspect, the present application provides a self-learning virtual display system based on external stimulus input, which adopts the following technical solution:

[0125] A self-learning virtual display system based on external stimulus input, wherein a program of any one of the above self-learning virtual display methods runs in the self-learning virtual display system.

[0126] The self-learning virtual display system includes:

[0127] The data acquisition module is used to monitor and obtain the game process status and the external stimulus input data after normalization and quantification in real time;

[0128] A data generation module, used to generate corresponding original image data according to the game process status and external stimulus input data;

[0129] A data selection processing module is used to configure a selection box, and select the sampled image data by selecting the original image data through the selection box;

[0130] A data sending module is used to input the sampled image data into the trained mode recommendation model to obtain the classification result output by the mode recommendation model; the mode recommendation model is a classification neural network model after training;

[0131] The terminal auxiliary configuration module is used to configure the game virtual mode according to the classification results;

[0132] Among them, the external stimulus input data includes user physiological data, environmental parameter data and user operation data; the game process status includes plot status, free status and combat status; the game virtual mode includes NPC interaction and plot branching strategies, virtual environment simulation strategies and enemy behavior patterns and strategies; the original image data includes n image partitions with sorting rules, and the image partitions are correspondingly associated with the data categories of the external stimulus input data. The sorting rules represent the order of the weights of the data categories of the external stimulus input that affect the classification results.

[0133] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A self-learning virtual display method based on external stimulus input, characterized in that: include: S1, real-time monitoring to obtain the game process status and the external stimulus input data after normalization and quantization processing; S2, generating corresponding original image data according to the game process state and the external stimulus input data; S3, configuring a selection box to select the original image data to obtain sampled image data; S4, inputting the sampled image data into the trained mode recommendation model to obtain the classification result output by the mode recommendation model; S5, configuring the game virtual mode according to the classification result; The game process state includes plot state, free state and combat state; the external stimulus input data includes user physiological data, environmental parameter data and user operation data; the game virtual mode includes NPC interaction and plot branch strategy, virtual environment simulation strategy and enemy behavior mode and strategy; the original image data includes n image partitions with sorting rules, and the image partitions are correspondingly associated with the data categories of the external stimulus input data; The step S2 comprises: S21, configuring a blank picture of a specified size; S22, dividing the blank picture into a process identification area and a data identification area; S23, dividing the data identification area into n first-level image partitions; S24, dividing each of the first-level picture partitions into corresponding second-level picture partitions; S25, filling the external stimulus input data into the secondary image partition according to the subdivided categories, and when the filling progress reaches a preset progress threshold, proceeding to step S26; S26, sorting the n first-level image partitions according to the sorting rule; The filling pixels of different primary image partitions have different hues, the filling pixels of the secondary image partitions of the same primary image partition have the same hue but different color depths, and the numerical value of the subdivided category data is represented by the brightness of the filling pixels; The sorting rules include: Calculate the subdivided data fluctuation variance V corresponding to the subdivided category data of each data category of the external stimulus input data ij , and according to the segmented data fluctuation variance V ij Calculate the average fluctuation variance V corresponding to each data category of the external stimulus input data i ; Wherein, i is the number corresponding to the first-level image partition, and j is the number of the second-level image partition corresponding to the subdivided category data; According to the average volatility variance V i Sort the first-level image partitions according to the variance V of the segmented data fluctuations ij The secondary picture partitions are sorted within the primary picture partitions.

2. The self-learning virtual display method based on external stimulus input according to claim 1 is characterized in that: The breakdown data volatility variance V ij and the average volatility variance V i The methods for obtaining include: Among them, x ijk is the pixel brightness value of the k-th pixel in the j-th secondary image partition of the ith primary image partition, is the average brightness of pixels in the jth secondary image partition of the ith primary image partition, m ij is the total number of padded pixels in the jth secondary image partition of the ith primary image partition, p i is the total number of secondary image partitions of the i-th primary image partition.

3. The self-learning virtual display method based on external stimulus input according to claim 1 is characterized in that: The method for configuring the selection box includes: Calculate and obtain the equipment load assessment factor U; According to the equipment load evaluation coefficient U and the comprehensive fluctuation variance V, the selection area S, the selection content overlap C and the selection frequency f of the selection box are configured; Among them, the method for obtaining the comprehensive volatility variance V includes: The equipment load evaluation coefficient U is negatively correlated with the cut-off area S, the cut-off content overlap C, and the cut-off frequency f; The comprehensive fluctuation variance V is positively correlated with the cut-off area S, the overlap degree C of the cut-off content, and the cut-off frequency f.

4. The self-learning virtual display method based on external stimulus input according to claim 3 is characterized in that: The method for obtaining the equipment load evaluation coefficient U includes: Obtain device-related operating parameters, including CPU occupancy rate R1, GPU occupancy rate R2, memory usage rate R3, game refresh rate pressure value R4, and key hardware position temperature pressure value R5; Among them, X fMAX is the maximum supported refresh rate, X f(t) is the average refresh rate of the current t-th unit period, X fMAX and X fMIN They are the historical maximum refresh rate and historical minimum refresh rate in the same game process state, X W(t) and X W(t-1) are the average temperature of the current t-th unit period and the average temperature of the t-1-th unit period, respectively. WMAX and X WMIN are the historical maximum key position temperature and the historical minimum key position temperature in the same game process state, Kf is the preset compensation frame number, R 4SD and R 5SD All are preset pressure constants.

5. The self-learning virtual display method based on external stimulus input according to claim 4 is characterized in that: The method for configuring the selection area S, the selection content overlap C and the selection frequency f of the selection box includes: The configuration method of the cut-off area S includes: The configuration method of the overlap degree C of the selected content includes: The configuration method of the cut-off frequency f includes: Among them, S Max and S Min are the maximum and minimum cross-sectional areas, respectively, and C Max and C Min are the maximum overlap of selected content and the minimum overlap of selected content, respectively, and f Max and f Min are the maximum and minimum cut-off frequencies, μ1, μ2, and μ3 are preset adjustment coefficients, S∈[S Min ,S Max ], C∈[C Min ,C Max ], f∈[f Min ,f Max ],V SD , U SD They are respectively a preset variance threshold and a preset load critical threshold.

6. The self-learning virtual display method based on external stimulus input according to claim 5, characterized in that: The sampled image data includes at least one sampled data unit, and each sampled data unit corresponds to the primary image data selected by the selection box.

7. The self-learning virtual display method based on external stimulus input according to claim 1 is characterized in that: When the game process state is switched, a weak reminder setting inquiry box of the game virtual mode is sent to the user; Adjusting the virtual mode of the game according to the user's intention and setting result; The mode recommendation model is supplementarily trained according to the intention and the setting result, and the mode recommendation model after the supplementary training is uniquely associated with the user.

8. A self-learning virtual display system based on external stimulus input, characterized in that: Running the self-learning virtual display method based on external stimulus input as described in any one of claims 1 to 7 comprises: The data acquisition module is used to monitor and obtain the game process status and the external stimulus input data after normalization and quantification in real time; A data generation module, used to generate corresponding original image data according to the game process state and the external stimulus input data; A data selection processing module, used for configuring a selection box, and obtaining sampled image data by selecting the original image data through the selection box; A data sending module, used for inputting the sampled image data into a trained mode recommendation model to obtain a classification result output by the mode recommendation model; the mode recommendation model is a classification neural network model after training; A terminal auxiliary configuration module, used to configure the game virtual mode according to the classification result; The external stimulus input data includes user physiological data, environmental parameter data and user operation data; the game process state includes plot state, free state and combat state; the game virtual mode includes NPC interaction and plot branch strategy, virtual environment simulation strategy and enemy behavior mode and strategy; the original image data includes n image partitions with sorting rules, and the image partitions are correspondingly associated with the data categories of the external stimulus input data; The data generation module is also used to execute the steps including: S21, configuring a blank picture of a specified size; S22, dividing the blank picture into a process identification area and a data identification area; S23, dividing the data identification area into n first-level image partitions; S24, dividing each of the first-level picture partitions into corresponding second-level picture partitions; S25, filling the external stimulus input data into the secondary image partition according to the subdivided categories, and when the filling progress reaches a preset progress threshold, proceeding to step S26; S26, sorting the n first-level image partitions according to the sorting rule; The filling pixels of different primary image partitions have different hues, the filling pixels of the secondary image partitions of the same primary image partition have the same hue but different color depths, and the numerical value of the subdivided category data is represented by the brightness of the filling pixels; The sorting rules include: Calculate the subdivided data fluctuation variance V corresponding to the subdivided category data of each data category of the external stimulus input data ij , and according to the segmented data fluctuation variance V ij Calculate the average fluctuation variance V corresponding to each data category of the external stimulus input data i ; Wherein, i is the number corresponding to the first-level image partition, and j is the number of the second-level image partition corresponding to the subdivided category data; According to the average volatility variance V i Sort the first-level image partitions according to the variance V of the segmented data fluctuations ij The secondary picture partitions are sorted within the primary picture partitions.

Images