Live broadcast dynamic narration method based on end-cloud collaboration and live broadcast interaction system

Abstract

The invention discloses a live broadcast dynamic narration method and a live broadcast interaction system based on end-cloud collaboration, and the method comprises the steps: when a target event occurs in a live broadcast room, a cloud end determines the narration state of the target event according to user behavior data sent by a client; when the narrative state is changed, the cloud generates a target rendering instruction corresponding to the changed narrative state according to a preset state instruction mapping table; the cloud sends a target rendering instruction to the client; and the client performs audio-visual rendering on the live broadcast room interface in response to the target rendering instruction, so that an audio-visual effect matched with the changed narrative state is added to the live broadcast room interface. According to the method, the client is used for monitoring the user behavior data, the cloud is used for obtaining the narrative state of the target event from the user behavior data, the cloud is prevented from receiving a large amount of interaction event data, the load pressure of the server is effectively reduced, and the audio-visual effect of the live broadcast room interface is increased after the narrative state is updated. And the user experience and enthusiasm are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of network live broadcast, in particular to a live broadcast dynamic narration method based on end-cloud cooperation and a live broadcast interaction system. BACKGROUND

[0002] The core of the current network live broadcast PK or team confrontation is that users increase the points of the supported anchor by gifting virtual gifts or likes, and the points are displayed in the form of a simple progress bar or a number. However, the traditional point expression requires that all interaction data be uploaded to the server in real time for centralized calculation and processing, which causes great pressure on the server in a high-concurrency scenario and may cause delays, affecting user experience. SUMMARY

[0003] The present application provides a live broadcast dynamic narration method based on end-cloud cooperation and a live broadcast interaction system, aiming to.

[0004] In order to achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0005] A live broadcast dynamic narration method based on end-cloud cooperation, applied to a live broadcast interaction system comprising a client and a cloud, the method comprising:

[0006] When a target event occurs in a live broadcast room, the cloud determines the narration state of the target event according to user behavior data sent by the client; the target event includes a confrontation between multiple anchors; the user behavior data at least includes user ID, home team, interaction value index, and / or key event; the interaction value index is calculated by the client according to the interaction event of the user in the live broadcast room, and is used to quantify the enthusiasm of the user participating in the target event;

[0007] When the narration state changes, the cloud generates a target rendering instruction corresponding to the changed narration state according to a preset state instruction mapping table; the state instruction mapping table includes multiple sample narration states and corresponding rendering instructions;

[0008] The cloud sends the target rendering instruction to the client;

[0009] The client responds to the target rendering instruction to perform audio-visual rendering on the live broadcast room interface, so that the live broadcast room interface increases audio-visual effects matching the changed narration state.

[0010] Optionally, the process of calculating the interaction value index by the client according to the interaction event of the user in the live broadcast room comprises:

[0011] The client determines at least one interactive event of the user in the live room in a monitoring time window and corresponding indexes, basic weights, and occurrence time stamps;

[0012] The client substitutes the indexes, basic weights, and occurrence time stamps of the plurality of interactive events in the monitoring time window into a time-series interactive intensity algorithm model to calculate an interactive value index corresponding to the monitoring time window; the time-series interactive intensity algorithm model is , wherein represents the interactive value index, represents a sending time stamp of the sampled user behavior data of the monitoring time window, represents a total number of interactive events, represents an index of an interactive event, represents a basic weight of an interactive event, represents a decay constant, represents an occurrence time stamp of an interactive event.

[0013] Optionally, when a target event occurs in the live room, the cloud determines a narrative state of the target event according to the user behavior data sent by the client, and the method comprises the following steps.

[0014] When the live room starts to have the target event, the cloud initializes a narrative state machine; the narrative state machine performs logical judgment on battle features of the target event based on a preset state migration rule to output a corresponding narrative state;

[0015] Before the target event ends, the client monitors user behavior data of the user in the live room in real time;

[0016] The client sends the user behavior data to the cloud;

[0017] The cloud determines a current battle feature of the target event based on the user behavior data;

[0018] The cloud inputs the current battle feature into the narrative state machine to obtain an output result of the narrative state machine; the output result is used to represent a current narrative state of the target event.

[0019] Optionally, the method further comprises the following steps.

[0020] When the live room starts to have the target event, the cloud sends an initial rendering instruction to the client;

[0021] The client performs audio-visual rendering on a live room interface in response to the initial rendering instruction, so that the live room interface has a preset audio-visual effect; the preset audio-visual effect is used to create a pre-battle atmosphere of the target event.

[0022] Optionally, the cloud server determines the current situation feature of the target event based on the user behavior data, including:

[0023] The cloud server determines the user behavior data received from the plurality of client devices within a preset time window;

[0024] The cloud server groups the plurality of user behavior data according to the home team to obtain a data group corresponding to different teams participating in the target event;

[0025] The cloud server obtains a multi-dimensional feature of the different teams based on the data group corresponding to the different teams;

[0026] The cloud server generates the current situation feature of the target event based on the multi-dimensional feature of the different teams.

[0027] Optionally, the multi-dimensional feature includes one or more of total intensity, participation breadth, average intensity, fire concentration, and key event marker;

[0028] The total intensity includes a sum of all interaction value indicators in the corresponding data group;

[0029] The participation breadth includes a sum of non-repeated user IDs in the corresponding data group;

[0030] The average intensity includes a ratio of the total intensity to the participation breadth;

[0031] The fire concentration includes a variance of all interaction value indicators in the corresponding data group;

[0032] The key event marker is used to represent the presence of the key event in the corresponding data group.

[0033] Optionally, the narrative state machine is specifically configured to: if the total intensity of the different teams is less than a specified threshold value, and the participation breadth of any team is not empty, output the current narrative state as confrontation; if the total intensity of any team is greater than a first threshold value, output the current narrative state as conflict; if the total intensity of the different teams is greater than a second threshold value, and the difference between the total intensities of the different teams is less than a third threshold value, and / or any team has the key event marker, output the current narrative state as climax; if the change amount of the total intensity of a target team within a specified time is greater than a third threshold value, output the current narrative state as reversal, the target team being a team with the highest total intensity among the different teams; and when the target event enters the countdown time, output the current narrative state as the final.

[0034] Optionally, when the narrative state changes, the cloud generates a target rendering instruction corresponding to the changed narrative state according to a preset state instruction mapping table, including:

[0035] The cloud listens to a plurality of output results generated by the narrative state machine in real time;

[0036] If the current output result is different from the previous output result, the cloud determines that the narrative state changes;

[0037] The cloud determines the changed narrative state based on the current output result;

[0038] The cloud generates a target rendering instruction corresponding to the changed narrative state according to a preset state instruction mapping table.

[0039] A live interactive system, comprising:

[0040] A client and a cloud;

[0041] The client is configured to send user behavior data to the cloud when a target event occurs in a live room; the target event includes a confrontation between a plurality of anchors; the user behavior data at least includes a user ID, a home team, an interaction value index, and / or a key event; the interaction value index is calculated by the client according to the user's interaction events in the live room, and is used to quantify the user's enthusiasm for participating in the target event;

[0042] The cloud is configured to determine the narrative state of the target event according to the user behavior data sent by the client when the target event occurs in the live room; when the narrative state changes, a target rendering instruction corresponding to the changed narrative state is generated according to a preset state instruction mapping table; the state instruction mapping table includes a plurality of sample narrative states and corresponding rendering instructions; and the target rendering instruction is sent to the client;

[0043] The client is further configured to respond to the target rendering instruction to perform audio-visual rendering on the live room interface, so that the live room interface increases audio-visual effects matched with the changed narrative state.

[0044] Optionally, the client is specifically configured to:

[0045] Determine at least one interaction event of the user in the live room and the corresponding index, base weight, and occurrence timestamp in a monitoring time window;

[0046] The indexes, the basic weights, and the occurrence time stamps of the plurality of interaction events in the monitoring time window are substituted into a time sequence interaction intensity algorithm model to calculate an interaction value index corresponding to the monitoring time window. , wherein represents the interaction value index, represents a sending time stamp of the user behavior data sampled by the monitoring time window, represents the total number of interaction events, represents the index of the interaction event, represents the basic weight of the interaction event, represents a decay constant, represents the occurrence time stamp of the interaction event.

[0047] The technical scheme provided in the application determines the narrative state of the target event according to the user behavior data sent by the client when the target event occurs in the live room. When the narrative state changes, the cloud end generates a target rendering instruction corresponding to the changed narrative state according to a preset state instruction mapping table. The cloud end sends the target rendering instruction to the client. The client responds to the target rendering instruction to perform audio-visual rendering on the live room interface, so that the live room interface increases the audio-visual effect matched with the changed narrative state. The application monitors the user behavior data by using the client, and obtains the narrative state of the target event from the user behavior data by using the cloud end, thereby avoiding the cloud end from receiving a large amount of interaction event data, effectively reducing the load pressure of the server, and increasing the audio-visual effect of the live room interface after the narrative state is updated, thereby improving the user experience and enthusiasm. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0049] Figure 1 An architecture schematic diagram of a live interaction system provided by an embodiment of the application;

[0050] Figure 2 A system module interaction logic schematic diagram provided by an embodiment of the application;

[0051] Figure 3 A flowchart of a live dynamic narrative method based on end-cloud cooperation provided by an embodiment of the application;

[0052] Figure 4Another flowchart of a live dynamic narration method based on end-cloud collaboration provided by an embodiment of the present application is shown in FIG. 6.

[0053] Figure 5 A flowchart of a live dynamic narration method based on end-cloud collaboration provided by an embodiment of the present application is shown in FIG. 6.

[0054] Figure 6 A flowchart of a live dynamic narration method based on end-cloud collaboration provided by an embodiment of the present application is shown in FIG. 6.

[0055] Figure 7 An analysis logic diagram of a narration state machine provided by an embodiment of the present application is shown in FIG. 7. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0057] In the present application, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. The term “include”, “contain” or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement “including a…” does not exclude the presence of another identical element in the process, method, article or device including the element.

[0058] As shown in FIG. 1, an architecture diagram of a live interactive system provided by an embodiment of the present application is shown, which includes a client 100 and a cloud 200 to solve the problems of single-dimension interaction, lack of narration and large server pressure in the existing live PK. Figure 1

[0059] In some examples, the so-called single-dimension interaction means that the user interaction mode is single, lacks strategic depth and collaboration depth, and the experience tends to be the same.

[0060] In some examples, the so-called lack of narration means that the live PK process lacks story and plot development, and when the score gap is too large, the live content becomes dull, and the user's participation and willingness to pay decrease sharply.​

[0061] In some examples, the so-called server pressure refers to that all interaction data of the live PK scene needs to be uploaded to the server in real time for centralized calculation and processing, which causes great pressure on the server in a high concurrency scene and may produce delay, affecting user experience.

[0062] Optionally, the client is deployed with a real-time battle awareness module, which is a lightweight module running on a user terminal device (such as a mobile phone) and can capture user interaction events (such as rewards, comments, likes, etc.) in a live room in real time, preprocess and calculate the interaction events using a preset temporal interaction density (TID) algorithm model, obtain a standardized battle intensity index (i.e., an interaction value index), and upload the interaction value index and a small amount of key data (such as user ID, home team, and / or key events) as user behavior data to the cloud.

[0063] It should be noted that the live PK can be regarded as a target event, and the target event includes the confrontation between multiple anchors, each anchor corresponds to a team, and the team includes users participating in the target event through the client. Specifically, the number of teams participating in the live room target event is at least two.

[0064] In possible implementations, the real-time battle awareness module includes an interaction capture unit, a TID algorithm core, and a data reporting unit. The interaction capture unit can be used to capture user interaction events such as likes, swipes, gift giving, and barrage input in the live room. The TID algorithm core is a lightweight calculation unit for calculating the interaction value index corresponding to the interaction event using the TID algorithm model. The data reporting unit is used to package the calculation result of the TID algorithm model and a small amount of key data into user behavior data and send it to the cloud through a communication protocol.

[0065] Optionally, the cloud is deployed with a macro-narrative director module, which can be used to receive user behavior data sent from a large number of clients, but does not process massive raw interaction event logs, and runs a narrative state machine inside. The narrative state machine can logically determine the macro-narrative stage (such as beginning, development, climax, and final battle) of the current PK according to the summarized battle data, and issue global narrative instructions (equivalent to target rendering instructions) to all clients.

[0066] In possible implementations, the macro-narrative director module includes a data receiving and aggregating unit, a narrative state machine, and an instruction generating and broadcasting unit. The data receiving and aggregating unit is configured to receive and aggregate user behavior data sent by all clients, and to integrate battle features of a target event from all user behavior data. The narrative state machine is configured to determine a narrative state of the target event at any time according to the battle features determined at the time. The instruction generating and broadcasting unit is configured to send corresponding target rendering instructions to all clients when the narrative state changes.

[0067] Optionally, the client is also deployed with a dynamic visual rendering engine, which is configured to receive narrative instructions from the cloud, and to dynamically and nonlinearly change visual elements of the PK interface, such as scene landscape, building form, attack special effect, weather environment, etc., according to the instructions, so as to render the abstract confrontation process into a visual story with rich plots in real time.

[0068] In some examples, the interaction logic between the interaction capturing unit, the TID algorithm core, the data reporting unit, the data receiving and aggregating unit, the narrative state machine, the instruction generating and broadcasting unit, and the dynamic visual rendering engine can be referred to the interaction logic shown in Figure 2 .

[0069] Optionally, the live broadcast dynamic narrative method based on end-cloud collaboration implemented by the live broadcast interaction system can be referred to the process shown in Figure 3 , which includes the following steps.

[0070] S301: When a target event occurs in a live broadcast room, the cloud determines a narrative state of the target event according to user behavior data sent by a client.

[0071] The target event includes a confrontation between multiple anchors, and the user behavior data at least includes a user ID, a home team, an interaction value index, and / or a key event. The interaction value index is calculated by the client according to an interaction event of a user in the live broadcast room, and is used to quantify the enthusiasm of the user in participating in the target event.

[0072] In some examples, the key event includes, but is not limited to, an interaction event in which a user makes a large amount of consumption in the live broadcast room, such as a rocket in the live broadcast room.

[0073] Optionally, the process of calculating the interaction value index by the client according to the interaction event of the user in the live broadcast room can be referred to the steps shown in Figure 4 and the corresponding explanation and description.

[0074] It should be noted that the beginning of the live room to attract users is particularly important when the target event occurs, which can effectively stimulate the enthusiasm of users to participate in the target event, and accordingly, corresponding means are needed to improve the user experience when the target event begins to occur in the live room. In addition, with the continuous development of the target event until the end, corresponding means are also needed to maintain the enthusiasm of users participating in the target event. Generally speaking, the enthusiasm of users participating in the target event can be influenced by their own senses, such as vision and hearing, and accordingly, the audio-visual effect of the live room interface can be improved to stimulate the feelings of users participating in the target event, thereby improving the enthusiasm and experience of users participating in the target event. As for what kind of audio-visual effect to take, it needs to be determined based on the narrative state of the target event, and accordingly, the cloud needs to listen to the narrative state of the target event based on the user behavior data sent by each client.

[0075] Optionally, when the target event occurs in the live room, the cloud determines the implementation process of the narrative state of the target event according to the user behavior data sent by the client, which can be seen from Figure 5 and the corresponding explanation.

[0076] S302: When the narrative state changes, the cloud generates a target rendering instruction corresponding to the changed narrative state according to a preset state instruction mapping table.

[0077] Among them, the state instruction mapping table includes a plurality of sample narrative states and corresponding rendering instructions.

[0078] In some examples, the rendering instruction corresponding to each sample narrative state can be composed of one or more sets of visual performance instructions and / or auditory performance instructions.

[0079] In some examples, different sample narrative states correspond to different rendering instructions to trigger different UI rendering scenes of the live room interface. Assuming that different sample narrative states include prologue, confrontation, conflict, climax, reversal, and end, and the number of teams participating in the target event is 2, specifically, for the prologue, the UI rendering scene can be represented as: the live room interface shows that both bases are intact, and the scene is calm. For confrontation, the UI rendering scene can be represented as: the live room interface appears with a breeze, particle light effects, etc. For conflict, the UI rendering scene can be represented as: the live room interface PK score leading side begins to appear energy shield around the base, and the attack animation becomes obvious. For climax, the UI rendering scene can be represented as: the live room interface enters the "white-hot" mode, the scene weather becomes "thunderstorm", the building appears in a magnificent evolution form, and the attack special effect becomes magnificent. For reversal, the UI rendering scene can be represented as: the live room interface appears "miracle light" and other special visual effects to increase the visual momentum. For the end, the UI rendering scene can be represented as: the background music and overall visual rendering of the live room interface become extremely oppressive and epic.

[0080] S303: The cloud sends a target rendering instruction to the client.

[0081] The target rendering instruction is not only a fixed ID, but also can be parameterized. For example, the target rendering instruction generated by the cloud can be CMD_ID:BaseEvolve, params:{team:'A',level:3}, where level:3 is dynamically calculated according to the degree to which the total score of team A exceeds the threshold.

[0082] In some examples, the cloud also packages the target rendering instruction into a unified format data packet (such as a JSON object), and efficiently issues it to all clients at the same time through the broadcasting capability of the server.

[0083] S304: The client responds to the target rendering instruction to perform audio-visual rendering on the live room interface, so that the live room interface adds audio-visual effects matching the changed narrative state.

[0084] The client calls the dynamic visual rendering engine to analyze the target rendering instruction, smoothly changes the visual performance of the live room interface, realizes the evolution or mutation of the UI rendering scene, completes a narrative advance, and the rendering process is repeated until the target event ends.

[0085] The above-mentioned S301-S304 process uses the client to monitor user behavior data in the live room, uses the cloud to obtain the narrative state of the target event from the user behavior data, avoids the cloud from receiving a large amount of interactive event data, effectively reduces the load pressure of the server, and after the narrative state is updated, uses the cloud to generate a target rendering instruction and triggers the client to respond to the target rendering instruction, so that the live room interface adds audio-visual effects matching the changed narrative state, improves user experience and immersion, and stimulates user enthusiasm for participating in the target event, thereby effectively solving the problems of single-dimension interaction, lack of narrative, and high server pressure in existing live PK.

[0086] As shown in Figure 4 Another process schematic diagram of a live dynamic narrative method based on end-cloud collaboration provided by an embodiment of the present application is shown, which includes the following steps.

[0087] S401: The client determines at least one interactive event of the user in the live room, and the corresponding index, base weight, and occurrence timestamp in the monitoring time window.

[0088] The index of the interactive event can be determined based on the occurrence order of the interactive event in the monitoring time window. In addition, the base weight of the interactive event may be determined according to the type of the interactive event, for example, the = 0.1, a barrage of = 0.5, a gift of 10 gold coins = 10, etc.

[0089] It should be noted that the client collects the corresponding user behavior data by monitoring the time window to realize real-time monitoring of user behavior data.

[0090] In some examples, the duration corresponding to the monitoring time window can be set by the technician according to the actual situation, for example, the duration corresponding to the monitoring time window can be set to 1 second, then the client will send the user behavior data determined every 1 second to the cloud.

[0091] S402: The client substitutes the indexes, base weights, and occurrence time stamps of the plurality of interaction events in the monitoring time window into the time-series interaction intensity algorithm model to calculate the interaction value index corresponding to the monitoring time window.

[0092] Wherein, the time-series interaction intensity algorithm model is , in the formula represents the interaction value index, represents the sending time stamp of the sampled user behavior data in the monitoring time window, represents the total number of interaction events, represents the index of the interaction event, represents the base weight of the interaction event, represents the decay constant, represents the occurrence time stamp of the interaction event.

[0093] In some examples, controls the influence fading speed of the historical interaction event, the greater the value, the faster the influence of the historical interaction event decreases, and the more sensitive the interaction value index is to the current reaction, the value can be issued by the cloud to the client at the beginning of the target event to adapt to target events of different durations.

[0094] In possible implementations, when the live room starts to occur the target event, the cloud determines the corresponding target decay constant according to the duration of the target event, and sends the target decay constant to the client, so that the client updates the original decay constant in the time-series interaction intensity algorithm model.

[0095] It should be noted that the time-series interaction intensity algorithm model runs on the client, and only a simple accumulation and exponential operation is needed every time a new interaction event occurs, the calculation cost is extremely low, and it does not affect the smoothness of the client main thread. In addition, by using the time-series interaction intensity algorithm model, a quantitative index reflecting the user's instantaneous participation enthusiasm is calculated The interaction value index depends not only on the interaction event itself (such as the size of the gift), but also on the time point and frequency of the interaction event.

[0096] The above-mentioned S401-S402 flow calculates the interaction value index corresponding to multiple interaction events in the monitoring time window by running the time sequence interaction intensity algorithm model on the client, which is an important part of the user behavior data, provides effective data support for the cloud to monitor the narrative state of the target event, and can also reduce the computing pressure of the cloud.

[0097] As shown in Figure 5 Another flowchart of a live dynamic narrative method based on end-cloud collaboration provided by an embodiment of the present application is shown, which includes the following steps.

[0098] S501: When the target event starts in the live room, the cloud initializes the narrative state machine.

[0099] The narrative state machine logically judges the battle features of the target event based on the preset state transition rules to output the corresponding narrative state.

[0100] In some examples, the narrative state machine outputs the corresponding narrative state based on the battle features of the target event as input. The narrative state machine maintains a current narrative state of the target event inside, and logically judges the input battle features according to a set of preset state transition rules to output the corresponding narrative state.

[0101] In possible implementations, the expression statement of the state transition rule can be: IF current state is S1 AND (A team Itotal> threshold $\theta_1$ OR B team Itotal> threshold $\theta_1$) THEN trigger the transition to S2 state.

[0102] In possible implementations, the types of narrative states can include S0 state, S1 state, S2 state, S3 state, S4 state, S5 state, etc.

[0103] In possible implementations, in addition to the hard threshold rule, the state transition rule also includes a trend analysis rule, which can be used to analyze multiple consecutive battle features to judge the battle trend of the target event, for example, if the growth rate of Itotal of the lagging side in the target event is detected to be continuously greater than a certain slope for 3 seconds, even if the absolute threshold is not reached, S4 state can be triggered in advance, that is, the current narrative state output by the narrative state machine is S4 state.

[0104] In a possible implementation, when the narrative state machine detects that the battle feature meets the corresponding state transition rule, the current narrative state is updated to the corresponding target state, and the target state is output as the current narrative state of the target event.

[0105] In some examples, the types of narrative states can be defined as prologue, confrontation, conflict, climax, reversal, and denouement. When the target event starts in the live room, the initialized narrative state machine outputs the current narrative state of the target event as the prologue by default.

[0106] S502: The cloud sends an initial rendering instruction to the client.

[0107] When the target event starts in the live room, the narrative state machine outputs the current narrative state of the target event as the prologue by default, which can trigger the cloud to send an initial rendering instruction to the client.

[0108] In some examples, S501 and S502 can be executed concurrently.

[0109] S503: The client performs audio-visual rendering on the live room interface in response to the initial rendering instruction, so that the live room interface increases preset audio-visual effects.

[0110] The preset audio-visual effects are used to create a pre-battle atmosphere of the target event.

[0111] It can be understood that when the target event starts in the live room, the pre-battle atmosphere of the target event is created in the live room interface, which can effectively improve the enthusiasm and experience of users participating in the target event.

[0112] S504: The client monitors user behavior data of users in the live room in real time before the target event ends.

[0113] The client can call the real-time battle awareness module mentioned above to monitor the user behavior data of users in the live room in real time after the target event occurs until the target event ends.

[0114] S505: The client sends the user behavior data to the cloud.

[0115] The client sends new user behavior data to the cloud every time new user behavior data is monitored, until all user behavior data generated in the execution process of the target event is sent to the cloud.

[0116] S506: The cloud determines the current battle feature of the target event based on the user behavior data.

[0117] The cloud end receives user behavior data sent by multiple clients at a current time point, and sorts out current battle situation features corresponding to the current time point from the user behavior data sent by the multiple clients.

[0118] Optionally, the cloud end determines the current battle situation features of the target event based on the user behavior data. For details, refer to Figure 6 and corresponding explanations.

[0119] S507: The cloud end inputs the current battle situation features into a narrative state machine to obtain an output result of the narrative state machine.

[0120] The output result is used to represent a current narrative state of the target event.

[0121] In some examples, the narrative state machine predefines multiple levels of narrative states, such as prologue, confrontation, conflict, climax, reversal, and denouement, and combines multiple dimensions of features involved in the current battle situation features, such as total intensity, participation breadth, average intensity, fire concentration, and key event marker, so that the narrative state machine can implement various effective logical judgments to obtain the current narrative state of the target event.

[0122] Optionally, the narrative state machine is specifically used for: if the total intensity of different teams is less than a specified threshold value, and the participation breadth of any team is not empty, outputting the current narrative state as confrontation; if the total intensity of any team is greater than a first threshold value, outputting the current narrative state as conflict; if the total intensity of different teams is greater than a second threshold value, and the difference between the total intensities of different teams is less than a third threshold value, and / or any team has a key event marker, outputting the current narrative state as climax; if the total intensity of a target team changes by more than the third threshold value within a specified time, outputting the current narrative state as reversal, and the target team is a team with the highest total intensity among different teams; and when the target event enters a countdown time, outputting the current narrative state as denouement.

[0123] In some examples, the target team can be regarded as a streamer whose PK score lags behind the first place among multiple streamers participating in the target event.

[0124] In possible implementations, the analysis logic of the narrative state machine for the current battle situation features can refer to Figure 7 .

[0125] In some examples, in the process of logical judgment of the current battle situation characteristics by the narrative state machine, the total intensity can be considered as the core basis for judging whether the energy basis for upgrading the narrative state is reached, the combination of the participation breadth and the average intensity can be used to distinguish whether it is "elite-driven" or "human wave tactics", so as to trigger different visual effects (such as generating a special light effect for "hero moment" for elite users), and the fire concentration degree can be used to judge the fire distribution, if the fire concentration degree of the team is extremely large, the visual effect of highlighting "sharp soldiers" can be triggered, otherwise the visual effect of the "legion" attack animation can be triggered.

[0126] The cloud can obtain the current battle situation characteristics of the target event by using the user behavior data sent by the client, analyze the current battle situation characteristics by the narrative state machine, obtain the current narrative state of the target event, and provide effective data support for rendering the live room interface of the client, thereby creatively transforming the monotonous live PK into an interactive movie with ups and downs written by all participants.

[0127] As shown in Figure 6 Another flowchart of a live dynamic narrative method based on end-cloud cooperation provided by the embodiment of the application is shown, which includes the following steps.

[0128] S601: The cloud determines the user behavior data sent by multiple clients in a preset time window.

[0129] The cloud can collect all user behavior data occurring at the same time point in each client according to the preset time window.

[0130] In some examples, the duration corresponding to the preset time window can be set by technical personnel according to actual conditions.

[0131] S602: The cloud groups the multiple user behavior data according to the home team to obtain data groups corresponding to different teams participating in the target event.

[0132] The data group includes multiple user behavior data of the same team.

[0133] In some examples, the home team in the user behavior data can be understood as the anchor supported by the user in the target event.

[0134] S603: The cloud obtains multi-dimensional characteristics of different teams based on the data groups corresponding to different teams.

[0135] For different teams, the cloud can perform big data analysis on the user behavior data corresponding to the team to obtain multi-dimensional characteristics of different teams.

[0136] Optionally, multi-dimensional features include total intensity, and one or more of the following: breadth of participation, average intensity, concentration of firepower, and key event markers; where total intensity includes the sum of all interaction value indicators in the corresponding data set; breadth of participation includes the sum of unique user IDs in the corresponding data set; average intensity includes the ratio of total intensity to breadth of participation; concentration of firepower includes the variance of all interaction value indicators in the corresponding data set; and key event markers are used to characterize the presence of key events in the corresponding data set.

[0137] In some examples, total intensity can reflect the team's total energy output, participation breadth can reflect the size of the team's participants, average intensity can reflect the team's average firepower per person, and firepower concentration can reflect the uniformity of the team's firepower distribution.

[0138] S604: The cloud generates current battle situation characteristics of target events based on the multi-dimensional characteristics of different teams.

[0139] In this process, the cloud-based system structures the multi-dimensional characteristics of different teams to generate the current battle situation characteristics of the target event.

[0140] As shown in S601-S604 above, the cloud can use user behavior data sent by the client to obtain the current combat situation characteristics of the target event, so as to provide effective data support for determining the current narrative state of the target event.

[0141] While several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of this application. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0142] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A live streaming dynamic narrative method based on edge-cloud collaboration, characterized in that, Applied to a live interactive system that includes a client and a cloud, the method includes: When a target event occurs in the live stream, the cloud determines the narrative state of the target event based on user behavior data sent by the client; the target event includes a confrontation between multiple streamers; the user behavior data includes at least user ID, team affiliation, interaction value index, and / or key events; the interaction value index is calculated by the client based on the user's interaction events in the live stream and is used to quantify the user's enthusiasm for participating in the target event; When the narrative state changes, the cloud generates a target rendering instruction corresponding to the changed narrative state according to a preset state instruction mapping table; the state instruction mapping table includes multiple sample narrative states and their corresponding rendering instructions. The cloud sends the target rendering command to the client; The client responds to the target rendering command and performs audiovisual rendering on the live broadcast interface, thereby adding audiovisual effects to the live broadcast interface that match the changed narrative state.

2. The method according to claim 1, characterized in that, The process by which the client calculates the interaction value index based on user interaction events in the live stream includes: The client determines at least one interactive event of the user in the live broadcast room under the monitoring time window, as well as the corresponding index, basic weight, and occurrence timestamp; The client substitutes the indices, basic weights, and occurrence timestamps of multiple interactive events within the monitoring time window into the time-series interaction intensity algorithm model to calculate the interaction value index corresponding to the monitoring time window; the time-series interaction intensity algorithm model is... In the formula Represents interactive value metrics. The timestamp representing the sending of user behavior data sampled within the monitoring time window. Represents the total number of interactive events. An index representing interactive events. Represents the basic weight of interactive events. Represents the attenuation constant. Represents the timestamp of the interactive event.

3. The method according to claim 1, characterized in that, When a target event occurs in the live stream, the cloud determines the narrative state of the target event based on user behavior data sent by the client, including: When a target event begins to occur in the live broadcast room, the cloud initializes the narrative state machine; the narrative state machine performs logical judgments on the battle characteristics of the target event based on preset state transition rules, and outputs the corresponding narrative state; Before the target event ends, the client monitors the user's behavior data in the live broadcast room in real time; The client sends the user behavior data to the cloud. Based on the user behavior data, the cloud determines the current combat situation characteristics of the target event; The cloud inputs the current battle situation features into the narrative state machine to obtain the output of the narrative state machine; the output is used to characterize the current narrative state of the target event.

4. The method according to claim 3, characterized in that, The method further includes: When the target event begins to occur in the live broadcast room, the cloud sends an initial rendering command to the client; The client responds to the initial rendering command and performs audiovisual rendering on the live broadcast interface, thereby adding preset audiovisual effects to the live broadcast interface; the preset audiovisual effects are used to create a pre-battle atmosphere for the target event.

5. The method according to claim 3, characterized in that, Based on the user behavior data, the cloud determines the current combat situation characteristics of the target event, including: The cloud determines the user behavior data sent by multiple clients within a preset time window; The cloud platform groups multiple user behavior data according to their team affiliation to obtain data groups corresponding to different teams participating in the target event. The cloud platform obtains multi-dimensional characteristics of different teams based on data groups corresponding to different teams; The cloud platform generates the current battle situation characteristics of the target event based on the multi-dimensional characteristics of the different teams.

6. The method according to claim 5, characterized in that, The multidimensional features include total intensity, as well as one or more of the following: breadth of participation, average intensity, concentration of firepower, and key event markers. The total intensity includes the sum of all interaction value indicators in the corresponding data set; The breadth of participation includes the sum of unique user IDs in the corresponding data group; The average intensity includes the ratio of the total intensity to the breadth of participation; The concentration of firepower includes the variance of all interactive value indicators in the corresponding data set; The key event marker is used to indicate that the key event exists in the corresponding data group.

7. The method according to claim 6, characterized in that, The narrative state machine is specifically used to: output the current narrative state as confrontation if the total intensity of all different teams is less than a specified threshold and the participation breadth of any team is not empty; output the current narrative state as conflict if the total intensity of any team is greater than a first threshold; output the current narrative state as climax if the total intensity of all different teams is greater than a second threshold and the difference in total intensity between different teams is less than a third threshold, and / or any team has the key event marker. If the change in the total intensity of the target team within a specified time exceeds the third threshold, the current narrative state output is "Reversal". The target team is not the team with the highest total intensity among the different teams. When the target event enters the countdown time, the current narrative state output is "End".

8. The method according to claim 3, characterized in that, When the narrative state changes, the cloud generates a target rendering instruction corresponding to the changed narrative state based on a preset state instruction mapping table, including: The cloud monitors multiple output results generated in real time by the narrative state machine; If the current output is different from the previous output, the cloud determines that the narrative state has changed; The cloud platform determines the changed narrative state based on the current output result; The cloud platform generates target rendering instructions corresponding to the changed narrative state based on a preset state instruction mapping table.

9. A live interactive system, characterized in that, include: Client and cloud; The client is used to: send user behavior data to the cloud when a target event occurs in the live broadcast room; the target event includes a competition between multiple streamers; the user behavior data includes at least user ID, team affiliation, interaction value index, and / or key events; the interaction value index is calculated by the client based on the user's interaction events in the live broadcast room, and is used to quantify the user's enthusiasm for participating in the target event; The cloud is used to: determine the narrative state of the target event based on user behavior data sent by the client when the target event occurs in the live broadcast room; When the narrative state changes, a target rendering instruction corresponding to the changed narrative state is generated according to a preset state instruction mapping table; the state instruction mapping table includes multiple sample narrative states and corresponding rendering instructions; the target rendering instruction is sent to the client. The client is also configured to: respond to the target rendering instruction, perform audiovisual rendering on the live streaming interface, so that the live streaming interface adds audiovisual effects that match the changed narrative state.

10. The system according to claim 9, characterized in that, The client is specifically used for: Determine at least one interactive event of a user in the live broadcast room within the monitoring time window, as well as the corresponding index, basic weight, and occurrence timestamp; The indexes, basic weights, and occurrence timestamps of multiple interactive events under the monitoring time window are substituted into the time-series interaction intensity algorithm model to calculate the interaction value index corresponding to the monitoring time window. The temporal interaction intensity algorithm model is as follows: In the formula Represents interactive value metrics. The timestamp representing the sending of user behavior data sampled within the monitoring time window. Represents the total number of interactive events. An index representing interactive events. Represents the basic weight of interactive events. Represents the attenuation constant. Represents the timestamp of the interactive event.

Images