A multi-mode music game scoring method

Abstract

The application discloses a kind of multi-mode music game score production methods;It is related to music game development technical field, including the beat data of target music, game mode set and score editing data, generate the beat sequence with event association identifier;According to the event range determined by editing trigger position and editing object relationship data, generate the rhythm event for expressing the same rhythm under different game mode unified association;The mode event of corresponding game mode is converted to score editing data, and is packaged as candidate export score data;According to the internal analysis rule corresponding to game engine analysis rule, generate running score, based on running back connection result, judge the back connection state of running event, running beat attribution relationship and running mode type relationship, output standardized multi-mode score data or score link exception record.The method is favorable to improve the standardization export and running back connection verification capability of multi-mode score data.

Description

Technical Field

[0001] This invention relates to the field of music game development technology, and more specifically, to a method for creating music game charts in a multi-mode format. Background Technology

[0002] In animation and game production engine software and development systems, music game charts typically need to be edited based on the beat position of the target music, the timing of interactive triggers, and the performance rules of different game modes. Existing chart creation methods mostly focus on a single game mode, recording editable objects such as key presses, trajectories, long presses, swipes, or channel triggers through a human-computer interface, and exporting chart files according to the data structure of the corresponding mode. For multi-mode music games, chart objects differ in display form, judgment methods, and data fields across different game modes. The same musical rhythm position may be recorded as different types of editable objects in different modes. Existing methods usually generate and export chart data for each mode separately, lacking an event correlation relationship based on the target music beat as a unified benchmark. When the chart data enters the game engine's parsing stage, whether the beat position, mode events, and exported fields in the editing state are consistent often relies on manual inspection or discovering anomalies after running the game. It is difficult to promptly determine whether the running events in the exported chart can be reconnected to the original rhythm expression and its corresponding beat nodes.

[0003] In the process of creating music game charts using existing technologies, there is a problem that it is difficult to form standardized multi-mode chart data that can be parsed and verified based on a unified beat correlation, when chart editing data for different game modes is used. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for creating multi-mode music game charts. It unifies the association of chart editing data generated through human-computer interaction across different game modes via beat sequences, and performs runtime analysis and backlink verification on candidate exported chart data. This solves the problems of unifying the organization, standardizing the export, and verifying the consistency of existing multi-mode chart data. Furthermore, it incorporates reinforcement learning to optimize the mode event transition strategy.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for creating music charts for multi-mode music games, comprising: S1. Obtain the beat data, game mode set, and chart editing data of the target music. The chart editing data includes the current editing mode belonging to the game mode set, the editing trigger position, the type of editing object, and the relationship data of the editing objects. S2. Generate a beat sequence based on the beat data. The beat sequence includes multiple beat nodes arranged in the order of playback of the target music and with event association identifiers. S3. Based on the edit trigger position and the relationship data of the edit object, determine the range of the event in the beat sequence, and generate rhythm events for uniformly associating the same rhythm expression in different game modes. The rhythm events are associated with the beat nodes within the range of the event. S4. Based on the current editing mode and the type of the edited object, convert the score editing data into mode events corresponding to the game mode, and associate the mode events with the corresponding rhythm events. The mode events record the mode type, event performance relationship, and event judgment relationship. S5. Based on the rhythm events and the pattern events associated with the rhythm events, generate preview chart data for the corresponding game modes, and encapsulate the rhythm events, pattern events, and the relationship between the pattern events and the beat nodes into candidate exported chart data. S6. Parse the candidate exported chart data according to the internal parsing rules corresponding to the game engine parsing rules, generate a chart including running events, running beat attribution relationships, and running mode type relationships, map the chart to beat sequences, rhythm events, and mode events, obtain running backlink results, and determine whether running events are backlinked to rhythm events, running beat attribution relationships are backlinked to beat nodes, and running mode type relationships are backlinked to mode events based on the running backlink results, and output standardized multi-mode chart data or chart link anomaly records according to the judgment results.

[0006] Beneficial effects: 1. Using the beat sequence as a unified benchmark, rhythm events, pattern events, and running backlink results are continuously organized, so that the chart editing data formed in different game modes can enter the same association link, thereby relatively improving the standardization export, parsing, and backlink verification capabilities of multi-mode chart data. 2. First, collect the beat data, game mode set and score editing data according to the music identifier, and then convert the editing object relationship data into a unified relationship type mark and relationship position mark. This allows the trigger beat node and the range of the event in the beat sequence to be determined more stably, so that cross-mode editing objects can be more stably merged into the same rhythm event. 3. Generate mode events using mode conversion templates and object conversion entry points, and record mode type, event performance relationship and event judgment relationship in mode events, so that the display differences and judgment differences of different game modes can be retained with mode events, while not destroying the unified association at the rhythm event level; 4. Write the rhythm event identifier, pattern event identifier, beat node identifier and pattern type into the common field area, and write the event performance relationship and event judgment relationship into the pattern field area. This allows the candidate exported spectrum data to still be linked back to the edit state object by means of the field source mapping record after internal parsing, thereby relatively reducing the risk of distortion of the association relationship after parsing. 5. This invention determines the spectrum object to be corrected based on the spectrum link anomaly record, and transforms the correction process into enhanced learning state data, enhanced learning action data and reward data. Then, the effectiveness of the strategy is verified by comparing the spectrum verification results before and after the update, so as to reduce the spectrum link anomalies in subsequent similar spectrum editing data. Attached Figure Description

[0007] Figure 1 This is a flowchart of the present invention; Figure 2 This is a flowchart illustrating the generation of beat sequences, determination of event range, and generation of rhythmic events in this invention. Figure 3 A flowchart for generating the pattern event conversion and the relationship between pattern events and rhythm events in this invention; Figure 4 This is a schematic diagram illustrating the relationship between the fields of the preview spectrum data generation and the candidate exported spectrum data in this invention. Figure 5 This is a flowchart of the process for internal parsing of candidate derived spectral data, running back connection judgment, and spectral output in this invention. Figure 6 This is a flowchart illustrating the spectral link anomaly correction, reinforcement learning training record generation, and mode event conversion strategy update verification process of the present invention. Detailed Implementation

[0008] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0009] Refer to the instruction manual appendix Figures 1 to 6 This paper describes a method for creating multi-mode music game charts. The implementation uses a multi-mode music game chart creation scenario within an animation / game engine software as an example. In this scenario, the same target music needs to be presented in bubble mode, track mode, and rhythm mode charts. Editors input or adjust chart editing data for different game modes through a human-computer interaction interface. Using the beat data of the target music as a unified benchmark, editing objects in different game modes are associated with beat sequences, rhythm events, and mode events, generating previewable and exportable candidate export chart data. Subsequently, running charts are generated according to internal parsing rules corresponding to the game engine's parsing rules, and the back-link relationship between the running charts and the edited data is verified. When chart link anomalies exist, anomaly records and correction records are generated. Furthermore, the mode event conversion strategy is updated through reinforcement learning.

[0010] S1. Obtain the beat data, game mode set, and chart editing data of the target music. The chart editing data includes the current editing mode belonging to the game mode set, the editing trigger position, the type of editing object, and the relationship data of the editing objects. Refer to the instruction manual appendix Figure 1 This implementation process establishes data attribution and object relationship expressions for the multi-mode score production data of the target music before it enters the beat sequence processing. In the animation and game production engine software, the beat data, game mode set, and score editing data come from the target music resources, mode configuration, and human-computer interaction editing records, respectively. The score object structure has field differences in different game modes. If the same target music is not collected and the relationship mark conversion is not completed first, it will be difficult to stably determine the trigger beat node, the set of relationship extension nodes, and the scope of the event. This implementation process first collects the beat data, game mode set, and score editing data with music identifiers to form the editing dataset to be processed. Then, it identifies the mode structure type corresponding to the current editing mode. Finally, it converts the editing object relationship data into a unified relationship type mark and relationship position mark to obtain the editing record to be processed.

[0011] Using the music identifier corresponding to the target music as an index, beat data, game mode sets, and chart editing data are grouped under the same target music to form a dataset to be processed. The music identifier is read from the target music resource record, chart project file, or music project configuration to distinguish the data belonging to different target music. Beat data is read from the chart project file, beat configuration record, or music project configuration corresponding to the target music, including the beat tempo, beat start time, and beat end time required to generate the subsequent beat sequence. The game mode set is read from the mode configuration record of the chart project, indicating the range of game modes that the target music needs to create or support. Chart editing data is read from the editing operation records generated by the editor through the human-computer interaction interface, including the current editing mode belonging to the game mode set, the editing trigger position, etc. The data includes the type of editable object and the relationship data between editable objects. The current editing mode indicates the game mode to which this editing operation belongs, the editing trigger position indicates the position of the editing operation on the target music playback timeline or beatline, the type of editable object indicates the category of the score object formed by this editing, and the relationship data between editable objects indicates the start point, end point, continuity, combination, synchronization, or persistence relationship between editable objects. During the aggregation process, music identifiers are used as search conditions to extract the beat records corresponding to the target music from the beat data, the mode configuration records corresponding to the target music from the game mode set, and the editing operation records corresponding to the target music from the score editing data. The beat data, game mode set, and score editing data belonging to the same music identifier are bound into a dataset to be processed, so that subsequent pattern structure recognition and editable object relationship transformation are all performed around the same target music.

[0012] The process involves identifying the chart object structure for each game mode in the game mode set to determine the mode structure type corresponding to the current editing mode. The chart object structure is read from the mode configuration file, chart object template, or mode editor configuration table corresponding to the game mode set, representing the data composition of editable chart objects in the corresponding game mode. The mode structure type indicates the organizational relationship between chart object fields in the current editing mode, used to interpret the relational and positional fields in the edited object's relational data. During processing, the current editing mode in the dataset to be processed is matched with the game mode identifier in the game mode set to determine the game mode to which this editing operation belongs. Then, the chart object structure corresponding to the game mode is read, and the structure of the chart object is identified. The system includes three fields: object composition, object position, and object relationship. The object composition field represents the basic data items contained in the score object; the object position field represents the position of the score object in the target music playback timeline, beatline, pathline, or channel position; and the object relationship field represents the connection, continuity, combination, or synchronization relationship between score objects. Based on the organization of the object composition, object position, and object relationship fields, the mode structure type corresponding to the current editing mode is determined. In this implementation, the bubble mode corresponds to a single-point trigger position structure, the trajectory mode corresponds to a path node connection structure, and the beat mode corresponds to a channel position and continuity range structure. This provides a clear field basis for subsequently reading the relationship category and relationship position from the edited object relationship data.

[0013] According to the pattern structure type corresponding to the current editing mode, the edit object relationship data in the score editing data is converted into unified relationship type markers and relationship position markers, resulting in the edit records to be processed for the beat sequence. Unified relationship type markers characterize the relationship categories in the edit object relationship data, and relationship position markers characterize the relationship positions in the edit object relationship data; both types of markers are derived from the edit object relationship data. During the conversion process, score editing data is read from the edit dataset to be processed, and the current editing mode, edit trigger position, edit object type, and edit object relationship data are extracted. Then, according to the pattern structure type corresponding to the current editing mode, the fields representing relationship categories and the fields representing relationship positions in the edit object relationship data are located. Fields representing relationship categories are converted into unified relationship type markers, which include point trigger, continuous extension, channel persistence, or synchronous combination. Fields representing relationship positions are converted into relationship positions. The markers, including trigger positions, extension positions, continuous positions, or channel positions, are used to define the relationship positions. In this implementation, the trigger relationship of clicked objects in bubble mode is converted into a unified relationship type marker for point triggers, and the position of the clicked object on the playback timeline or beat axis is converted into a relationship position marker for trigger positions. In trajectory mode, the connection relationship between path nodes is converted into a unified relationship type marker for continuous extensions, and the path start and end points are converted into relationship position markers for extension positions. In beat mode, the relationship between button objects and channels and their duration is converted into a unified relationship type marker for channel duration, and the channel position, duration start point, and duration end point are converted into relationship position markers for channel positions or duration positions. The pending edit record retains the current edit mode, edit trigger position, and edit object type, and carries the unified relationship type marker and relationship position marker as input for subsequently determining the trigger beat node, the set of relationship extension nodes, and the range of the event in the beat sequence.

[0014] S2. Generate a beat sequence based on the beat data. The beat sequence includes multiple beat nodes arranged in the order of playback of the target music and with event association identifiers. Refer to the instruction manual appendix Figure 1 and Figure 2 This implementation process converts the beat data of the target music into a beat sequence that can carry rhythm event relationships. In the production of multi-mode music game charts, although the chart objects in different game modes have different forms of expression, they all need to fall into the corresponding beat range according to the playback order of the target music. Therefore, it is necessary to first determine the beat division scale based on the beat speed, beat start time and beat end time, and then generate beat nodes with beat node identifiers, node time ranges and event association identifiers, so that the subsequent editing trigger position can match the beat node and the rhythm event identifier can be written into the corresponding beat node.

[0015] Based on the beat speed, beat start time, and beat end time in the beat data, the beat time interval between adjacent beat nodes and the number of beat nodes corresponding to the target music are calculated. The beat data comes from the beat records in the aforementioned dataset to be processed. The beat speed represents the number of beats in the target music per unit time. The beat start time represents the start playback time when the target music enters the score production range, and the beat end time represents the end playback time when the target music leaves the score production range. In the implementation method where the beat speed is recorded in beat configuration records in beats per minute, the beat time interval is obtained by dividing 60 seconds by the beat speed. The number of beat nodes corresponding to the target music is obtained by dividing the length of time between the beat end time and the beat start time by the beat time interval. If there is a remainder in the calculation result, it is rounded up to ensure that the beat nodes cover the complete score production range from the beat start time to the beat end time. The beat time interval serves as the time span for generating adjacent beat nodes, and the number of beat nodes serves as the basis for the number of beat nodes generated subsequently.

[0016] Generate beat nodes corresponding to the number of beat nodes according to the playback order of the target music, and configure a beat node identifier and node time range for each beat node; a beat node represents a beat unit divided according to the beat time interval on the playback timeline of the target music, the beat node identifier represents the unique number of the beat node in the target music, and the node time range represents the start and end boundaries of the beat node on the playback timeline of the target music; during the generation process, the start time of the beat is used as the starting point of the first beat node, and subsequent beat nodes are generated sequentially according to the beat time interval along the playback order of the target music; the nth beat node... The starting point of a node is obtained by adding n-1 beat intervals to the start time of the beat. The ending point of the nth beat node is obtained by adding the beat interval to the starting point of the nth beat node. When the ending point of the last beat node exceeds the end time of the beat, the end time of the beat is used as the end boundary of the time range of the last node. The beat node identifier is generated by combining the music identifier corresponding to the target music with the node order, or by generating it according to the numbering rules that can uniquely locate the beat node in the score engineering. The beat nodes formed in this way provide time and identifier references for subsequent editing trigger position matching and running beat belonging back connection.

[0017] Configure event association identifiers for each beat node to record rhythm event identifiers, generating a beat sequence. The event association identifier represents a reserved rhythm event association field in the beat node, used to write the corresponding rhythm event identifier after subsequent generation or matching of rhythm events. During configuration, the beat node identifier, node time range, and event association identifier are written to the same beat node record. The event association identifier remains unwritten before the rhythm event is generated, and is written to the corresponding rhythm event identifier by the rhythm event generation step after the range of the subsequent event is determined. Multiple beat node records are arranged according to the playback order of the target music to form a beat sequence. The beat sequence includes multiple beat nodes arranged according to the playback order of the target music and having event association identifiers, serving as the unified beat basis for subsequent triggering of beat node determination, relationship extension node set determination, rhythm event generation, and running score back-link verification.

[0018] S3. Based on the edit trigger position and the relationship data of the edit object, determine the range of the event in the beat sequence, and generate rhythm events for uniformly associating the same rhythm expression in different game modes. The rhythm events are associated with the beat nodes within the range of the event. Refer to the instruction manual appendix Figure 1 and Figure 2 This implementation process locates the score editing data within the beat node range of the target music and generates rhythm events that can carry the same rhythm expression in different game modes. Since the edited objects in Bubble Mode, Track Mode, and Rhythm Mode have different relationship forms such as click, path connection, and channel continuity, directly generating events according to the original fields of each mode would lead to difficulties in unifying the rhythm relationship across modes. Therefore, this implementation process first uses the editing trigger position to determine the trigger beat node, then determines the set of relationship extension nodes and the range of the event based on the relationship data of the edited object, and finally determines whether to reuse existing rhythm events or generate new rhythm events based on the event association identifier of the beat nodes within the event range, so that subsequent mode event conversions can be executed around the same rhythm event.

[0019] The editing trigger position in the score editing data is matched with the node time range of each beat node in the beat sequence to determine the trigger beat node. The editing trigger position comes from the aforementioned edit record to be processed and is used to indicate the position of this editing operation on the target music playback time axis or beat axis. The node time range comes from the beat nodes in the beat sequence and is used to indicate the start and end boundaries of the corresponding beat node on the target music playback time axis. During the matching process, the node time range of each beat node is read according to the playback order of the target music, and the editing trigger position is compared with the node time range one by one. When the editing trigger position is greater than or equal to a certain beat, the trigger is determined. When the start point of a time range is less than the end point of that time range, the beat node corresponding to that time range is determined as the trigger beat node. When the edit trigger position is equal to the end point of the time range of the last beat node, the last beat node is determined as the trigger beat node. When the edit trigger position is recorded as a beat axis position, it is first converted into the position on the target music playback time axis based on the beat start time and beat time interval, and then the node time range matching is performed. The trigger beat node indicates the starting beat of the score editing data in the beat sequence and serves as the reference node for subsequently determining the extended node set of relationships.

[0020] Based on the triggering beat node, the set of relationship extension nodes is determined in the beat sequence according to the relationship data of the edited object, and the triggering beat node and the set of relationship extension nodes are defined as the scope of the event. The set of relationship extension nodes represents the set of beat nodes that extend, continue, or are associated with the edited object relationship from the triggering beat node in the beat sequence, and the scope of the event represents the complete range of beat nodes that the rhythmic event needs to be associated with. The unified relationship type marker and relationship position marker carried in the edit record to be processed are converted from the edited object relationship data. The unified relationship type marker is used to characterize the relationship category in the edited object relationship data, and the relationship position marker is used to characterize the relationship position in the edited object relationship data. When determining the set of relationship extension nodes, the node selection rules are determined according to the unified relationship type marker, and the corresponding beat nodes are matched in the beat sequence according to the relationship position marker. When the unified relationship type marker indicates a point-triggered relationship, the set of relationship extension nodes is empty, and the scope of the event is composed of the triggering beat node. The unified relationship type marker indicates a continuous relationship. When extending a relationship, starting from the trigger beat node, the system searches along the target music playback sequence to find the beat node corresponding to the termination position in the relationship position marker. The beat nodes between the trigger beat node and the termination position are then defined as the relationship extension node set. When the unified relationship type marker indicates a channel continuity relationship, the system matches the corresponding node time range based on the continuity start and end points in the relationship position marker, and the beat nodes covered by the continuity range are defined as the relationship extension node set. When the unified relationship type marker indicates a synchronous combination relationship, the system reads the editing trigger position or relationship position marker corresponding to the combination object in the editing object relationship data, and matches this editing trigger position or relationship position marker with the node time range in the beat sequence. If the matched beat node is within the same node time range as the trigger beat node, or within the synchronous combination range defined by the editing object relationship data, the matched beat node is merged into the relationship extension node set. The event range is used to limit the association range between subsequent rhythm events and beat nodes.

[0021] Based on the event's scope and the relationship data of the edited objects, the event association identifiers of the beat nodes within the event's scope are matched. If a matching event association identifier exists, the chart editing data is associated with an existing rhythm event. If no matching event association identifier exists, a new rhythm event is generated, and the new rhythm event identifier is written to the beat nodes within the event's scope. The event association identifier is an association field in the beat node used to record the rhythm event identifier. The rhythm event represents an event object used to uniformly associate the same rhythm expression across different game modes. During the matching process, the event association identifiers of each beat node within the event's scope are read, and it is determined whether the event association identifier has already been written to the rhythm event identifier. If the event association identifier has already been written to the rhythm event identifier, the corresponding existing rhythm event is read, along with the event scope corresponding to the existing rhythm event and the time range of the nodes covered by the existing rhythm event. The event scope corresponding to the current chart editing data is the same as the event scope corresponding to the existing rhythm event, or the current chart editing data... If the corresponding event falls within the time range of nodes covered by existing rhythm events, and the trigger position, extension position, or duration position pointed to by the relationship position marker falls within the time range of nodes covered by existing rhythm events, the event association identifier is determined to match the current chart editing data, and the current chart editing data is associated with the existing rhythm events. The unified relationship type marker is associated with the existing rhythm events along with the current chart editing data, and is used to record the relationship category of the current chart editing data in the corresponding game mode. If no event association identifier is written within the event's range, or if the existing rhythm events corresponding to the events with written event association identifiers do not meet the above matching conditions with the current chart editing data, a new rhythm event is generated for the current chart editing data, a new rhythm event identifier is configured, and the new rhythm event identifier is written to the event association identifier of each beat node within the event's range. After writing, the rhythm event is associated with the beat nodes within the event's range, and a unified rhythm basis is provided for subsequent mode event conversions.

[0022] For example, near the same beat position of the same target music, bubble mode forms clickable objects, track mode forms path node objects that pass through the same beat range, and beat mode forms channel button objects that cover the same beat range. In this implementation process, the editing trigger position of the clickable object is first matched as the trigger beat node, and then the path node objects and channel button objects are determined according to the relationship position mark to determine the set of relationship extension nodes. If the event corresponding to these objects has already been written with a matching rhythm event identifier within its range, the corresponding score editing data is associated with the existing rhythm event. If there is no matching event association identifier within the event's range, a new rhythm event is generated and written to the corresponding beat node, so that the same rhythm expression in bubble mode, track mode, and beat mode can be transformed around the same rhythm event in subsequent steps.

[0023] S4. Based on the current editing mode and the type of the edited object, convert the score editing data into mode events corresponding to the game mode, and associate the mode events with the corresponding rhythm events. The mode events record the mode type, event performance relationship, and event judgment relationship. Refer to the instruction manual appendix Figure 1 and Figure 3 This implementation process converts the chart editing data into mode events that can be previewed, encapsulated, and parsed in the corresponding game mode. Since different game modes have different display objects and judgment rules for the same rhythm expression, bubble mode focuses on click triggers, track mode focuses on path connections, and rhythm mode focuses on channel buttons and continuous judgment. Therefore, it is necessary to first determine the conversion basis according to the current editing mode and the type of editing object, then generate mode events according to the conversion basis, and re-associate the mode events with the corresponding rhythm events and the beat nodes within the event range, so that subsequent preview chart data and candidate exported chart data can simultaneously retain cross-mode rhythm associations and the performance and judgment differences of each game mode.

[0024] Based on the current editing mode and the type of the edited object, a mode conversion template and an object conversion entry point are determined, generating a mode event conversion task. The current editing mode comes from the edit record to be processed and is used to determine the game mode to which the score editing data belongs. The edited object type comes from the score editing data and is used to represent the object category formed in this edit. The mode conversion template is the organization template for mode event fields under the corresponding game mode, derived from the mode conversion configuration in the animation and game production engine software, limiting the mode type, performance field, judgment field, and association field that need to be written for the mode event. The object conversion entry point is the entry rule for the edited object type to enter the corresponding conversion process, also derived from the mode conversion configuration, used to determine the conversion path corresponding to different edited object types. In the process, the current editing mode is first matched with the game mode identifier in the game mode set, and the mode conversion configuration of the matching game mode is read. Then, the object conversion entry is searched in the mode conversion configuration according to the editing object type, and the mode conversion template corresponding to the object conversion entry is read. The current editing mode, the editing object type, the mode conversion template, and the object conversion entry together constitute the mode event conversion task. In this implementation, the click object in the bubble mode matches the click event conversion entry, the path object in the trajectory mode matches the path event conversion entry, and the button object in the rhythm mode matches the channel button event conversion entry. This ensures that the subsequent conversion process can read and write the score editing data according to the data structure of the corresponding game mode.

[0025] The chart editing data is converted into mode events corresponding to the game mode according to the mode event conversion task. Mode event identifiers are configured for each mode event, and the event performance relationships and event judgment relationships of the mode events are calculated. Simultaneously, the mode conversion template, object conversion entry point, and converted mode events are recorded to obtain the mode event conversion record. A mode event represents a chart event that can participate in previewing, field encapsulation, and internal parsing in a certain game mode. The mode event identifier is used to uniquely identify the converted mode event. The event performance relationship indicates the display position, display path, display channel, duration range, or combined performance method of the mode event in the corresponding game mode. The event judgment relationship indicates the judgment time, judgment range, hit type, or continuous judgment range of the mode event in the corresponding game mode. During the conversion process, the current editing mode, editing trigger position, editing object type, editing object relationship data, unified relationship type marker, and relationship position marker are read from the chart editing data according to the object conversion entry point. Then, the read data is written into the corresponding mode event field according to the mode conversion template. The mode event identifier is generated according to the music identifier corresponding to the target music, the current editing mode, and the event sequence number. The item number is obtained by incrementing the number of generated mode events under the same music identifier and the same current editing mode, or by generating a numbering rule that can uniquely locate mode events in the score project; the event performance relationship is calculated based on the type of edited object and the relationship position mark. For example, the display position of a clicked object is formed based on the trigger position, the display path of a path object is formed based on the extension position, and the channel button object is formed based on the channel position and the continuous position to form the channel display range; the mode type is determined based on the current editing mode and written into the mode event to identify the game mode to which the mode event belongs; the event judgment relationship is calculated based on the editing trigger position, the range of the event, and the judgment field of the corresponding game mode. For example, the triggering beat node or triggering time is used as the judgment basis for a point triggering event, the time range of the node covered by the path is used as the judgment interval for a continuous extension event, and the time range of the node corresponding to the continuous start and continuous end point of a channel continuous event is used as the continuous judgment range; the mode event conversion record records the mode conversion template used this time, the object conversion entry, and the converted mode event, providing a traceable basis for the source of the mode field in the subsequent candidate exported score data and the generation of reinforcement learning action data.

[0026] The pattern event identifier is written to the corresponding rhythm event, and the pattern event is associated with the beat nodes within the event's range, generating an association record between the pattern event and the rhythm event. The corresponding rhythm event comes from the existing rhythm event or the newly generated rhythm event determined based on the event's range and the event association identifier. The beat nodes within the event's range include the trigger beat nodes and the beat nodes in the set of relationship extension nodes. During processing, the rhythm event identifier corresponding to the rhythm event is read first, and then the pattern event identifier is written to the pattern event association field of the rhythm event, enabling the rhythm event to record one or more corresponding pattern events. Subsequently, the pattern event identifier is associated with the beat node identifiers within the event's range, so that the pattern event is not only associated with the rhythm event but also corresponds to a specific beat node range. When the same rhythm event is already associated with pattern events in other game modes, the newly generated pattern event identifier is merged into the pattern event association field of the same rhythm event, without overwriting the existing pattern event identifier. The association record between the pattern event and the rhythm event includes at least the correspondence between the rhythm event identifier, the pattern event identifier, and the beat node identifier, and serves as the basis for the subsequent generation of preview chart data, fieldized chart encapsulation records, and candidate exported chart data.

[0027] For example, the same rhythmic position of the same target music forms a click object in bubble mode, a path node object in track mode, and a channel button object in rhythm mode. The animation and game production engine software determines the click event conversion entry, path event conversion entry, and channel button event conversion entry according to the current editing mode and the type of editing object, and generates three mode events according to the corresponding mode conversion template. The three mode events calculate the corresponding display position, display path or channel duration range, and corresponding judgment relationship, and then write the three mode event identifiers into the same rhythm event and establish an association with the beat node within the event range, so that the score events in bubble mode, track mode, and rhythm mode can enter the subsequent preview and export process around the same rhythm expression.

[0028] S5. Based on the rhythm events and the pattern events associated with the rhythm events, generate preview chart data for the corresponding game modes, and encapsulate the rhythm events, pattern events, and the relationship between the pattern events and the beat nodes into candidate exported chart data. Refer to the instruction manual appendix Figure 1 and Figure 4This implementation process is executed after the rhythm events, pattern events, and the association between pattern events and beat nodes have been established. The chart events of the corresponding game mode are first converted into preview chart data that can be viewed by editors. Then, they are encapsulated according to common association fields and pattern difference fields, and bound to the music identifier and editing version identifier corresponding to the target music, forming candidate exported chart data that can enter subsequent internal parsing and back-link verification. Since multi-mode music game charts need to retain cross-mode associations under the same rhythm event, as well as performance rules and judgment rules under different game modes, this implementation process uses three consecutive processes: preview generation, field partitioning and encapsulation, and candidate export binding, so that the subsequent parsing process can simultaneously read rhythm events, pattern events, and the association between pattern events and beat nodes.

[0029] Based on rhythm events, associated pattern events, and the beat nodes corresponding to the pattern events, preview chart data for the corresponding game mode is generated according to the current editing mode. Rhythm events originate from the aforementioned event range determination and rhythm event generation process; pattern events originate from the aforementioned pattern event conversion process; the beat nodes corresponding to the pattern events originate from the association between the pattern event and the beat nodes within its range; and the current editing mode originates from the edit records to be processed. During processing, pattern events corresponding to the game mode are first filtered from the pattern events associated with rhythm events according to the current editing mode. Then, the event performance relationships and event judgment relationships carried by the filtered pattern events are read, and the relationship between the pattern events and the pattern events is read. The preview chart data includes the associated beat node identifier and node time range; the event performance relationship is used to determine the display position, display path, display channel, duration range or combined display method of the chart object in the preview interface; the event judgment relationship is used to determine the judgment time, judgment interval, hit type or duration judgment range of the chart object in the preview interface; the node time range of the beat node is used to determine the display position of the chart object on the target music playback time axis or beat axis; the preview chart data generated thereby includes the mode events, event performance relationships, event judgment relationships and beat node association relationships in the current editing mode, and serves as the data basis for editors to view and correct the corresponding game mode chart content before candidate export.

[0030] The rhythm event identifier, pattern event identifier, beat node identifier, and pattern type are written into the common field area, and the event performance relationship and event judgment relationship are written into the pattern field area, forming a field-based chart encapsulation record. The common field area represents the associated field area read by different game modes, used to retain the common association relationship between rhythm events, pattern events, and beat nodes. The pattern field area represents the performance field and judgment field area read by the corresponding game mode, used to retain the event display rules and event judgment rules in the current editing mode. During processing, the rhythm event identifier in the rhythm event is read, the pattern event identifier and pattern type in the pattern event are read, and the beat node corresponding to the pattern event is read. The beat node identifier is then written to the beat field in the common field area; the mode event identifier is written to the mode event field in the common field area; the beat node identifier is written to the beat attribution field in the common field area; and the mode type is written to the mode type field in the common field area. The event performance relationship is written to the mode field area according to the display field format corresponding to the current editing mode, and the event judgment relationship is written to the mode field area according to the judgment field format corresponding to the current editing mode. The field-based chart encapsulation record is composed of the common field area and the mode field area, enabling subsequent internal parsing to first read the cross-mode common association fields, and then read the performance fields and judgment fields of the corresponding game mode.

[0031] The field-based chart encapsulation record is bound to the music identifier and editing version identifier corresponding to the target music to obtain candidate exported chart data. The music identifier comes from the target music resource record, chart project file, or music project configuration. The aforementioned data collection process has already used the music identifier to limit the beat data, game mode set, and chart editing data to the same target music. The editing version identifier is read from the editing record or export record of the current chart project and is used to distinguish the chart data formed in different editing stages under the same target music. During processing, the field-based chart encapsulation record is written to the candidate export record corresponding to the target music, and the music identifier and editing version identifier are written to the candidate export record at the same time. The field content in the common field area and the mode field area remains unchanged. When there are multiple field-based chart encapsulation records for the same game mode of the same target music, they are classified into the same target music according to the same music identifier, and the candidate exported chart data formed in different editing stages are distinguished according to the editing version identifier. The candidate exported chart data encapsulates rhythm events, mode events, and the relationship between mode events and beat nodes, which serve as inputs for subsequent internal parsing, runtime backlink judgment, and event association chain closure verification.

[0032] For example, when creating bubble mode, track mode, and rhythm mode charts for the same target music, the click object in bubble mode, the path object in track mode, and the channel button object in rhythm mode form the preview chart data for the corresponding game mode. During the encapsulation process, the rhythm event identifier, event identifiers for each mode, beat node identifier, and mode type corresponding to the same rhythm event are written into the common field area, and the click display position, path display track, channel display position, and corresponding judgment range are written into the mode field area. After binding the music identifier of the target music and the current editing version identifier, candidate exported chart data that can enter the subsequent parsing and verification process are formed.

[0033] S6. Parse the candidate exported chart data according to the internal parsing rules corresponding to the game engine parsing rules, generate a running chart including running events, running beat attribution relationships and running mode type relationships, map the running chart to beat sequences, rhythm events and mode events, obtain running backlink results, and determine whether the running events are backlinked to rhythm events, whether the running beat attribution relationships are backlinked to beat nodes and whether the running mode type relationships are backlinked to mode events based on the running backlink results, and output standardized multi-mode chart data or chart link anomaly records according to the judgment results; Refer to the instruction manual appendix Figure 1 , Figure 5 and Figure 6 After the candidate exported chart data is formed, it is verified whether the candidate exported chart data can form a runnable chart structure according to the internal parsing order corresponding to the game engine parsing rules, and further it is determined whether the run chart can be backlinked to the rhythm events, beat nodes and pattern events in the editing state. Although the candidate exported chart data has encapsulated the rhythm events, pattern events and the relationship between pattern events and beat nodes, it still needs to go through field integrity verification, internal parsing, run backlink judgment and event association chain closure verification before it can determine whether to output standardized multi-mode chart data or generate chart link anomaly records.

[0034] Candidate exported chart data first undergoes field integrity verification and internal parsing to determine if it possesses the basic field conditions for generating a running chart. Field integrity verification reads the common field area and pattern field area from the candidate exported chart data. The common field area includes rhythm event identifiers, pattern event identifiers, beat node identifiers, and pattern types. The pattern field area includes event performance relationships and event judgment relationships. During verification, it checks whether the rhythm event identifiers, pattern event identifiers, beat node identifiers, and pattern types exist in the common field area, whether the event performance relationships and event judgment relationships exist in the pattern field area, and whether a correspondence can be established between the event performance relationships and event judgment relationships in the pattern field area and the pattern event identifiers and pattern types in the common field area. If the field integrity verification passes, it proceeds according to the parsing rules of the game engine. The internal parsing process sequentially reads the common field area, the pattern field area, and the relationships between fields, parsing the candidate exported spectrum data into a running spectrum. During the parsing process, field source mapping records are retained. These records are used to record the correspondence between the running events in the running spectrum and the common field area and pattern field area in the candidate exported spectrum data. The running spectrum includes running events, running beat attribution relationships, and running mode type relationships. Running events represent the running spectrum events formed after internal parsing, running beat attribution relationships indicate the attribution of running events in the beat sequence, and running mode type relationships indicate the game mode to which the running events belong. When field integrity verification fails, a field integrity exception result is generated. This result records the location of missing fields, uncorresponding fields, or unparsable fields, and the internal parsing process is stopped.

[0035] After the running score is generated, the running events, running beat attribution relationships, and running mode type relationships are mapped to rhythm events, beat nodes, and mode events, respectively, thus forming the running backlink results. The mapping process reads the running event field from the running score. When the running event field directly carries a rhythm event identifier, the running event is mapped to the rhythm event in the edit state based on that identifier. When the running event field does not directly carry a rhythm event identifier, the field source mapping record retained during the internal parsing process is read, and the mapping is completed based on the rhythm event identifier corresponding to the common field area in the field source mapping record. The beat attribution field in the running beat attribution relationship is read, and the mapping is completed based on the beat attribution relationship. The node identifier or node time range maps the running beat attribution relationship to the beat node in the beat sequence; it reads the mode field in the running mode type relationship and maps the running mode type relationship to the mode event in the corresponding game mode according to the mode type and mode event identifier; the running back connection result records three types of mapping relationships, including the mapping relationship between running events and rhythm events, the mapping relationship between running beat attribution relationship and beat node, and the mapping relationship between running mode type relationship and mode event; through this mapping process, the running state object formed by the candidate exported score data after internal parsing can establish a verifiable correspondence with the previous edit state object.

[0036] The basic backlink judgment result is formed based on the running backlink result or the field integrity anomaly result. It is used to determine whether the running spectrum has the basic conditions to enter the event association chain closure verification. When the running backlink result is obtained, it is judged in turn whether the running event can be backlinked to the rhythm event, whether the running beat attribution relationship can be backlinked to the beat node, and whether the running mode type relationship can be backlinked to the mode event. When all three backlinks are true, a basic pass result is generated. When any backlink is false, a basic anomaly result is generated, and the object type, anomaly field position, and corresponding object identifier of the failed backlink are recorded. If a field integrity anomaly result has been generated, the field integrity anomaly result is written into the basic backlink judgment result, and the basic backlink judgment result is determined as the basic anomaly result. The basic backlink judgment result serves as the basis for subsequent branch judgment. The basic pass result allows the continued generation of event association chain verification records, while the basic anomaly result directly points to the spectrum link anomaly record generation process.

[0037] When the basic backlink judgment result is the basic pass result, an event association chain verification record is generated based on the running backlink result, rhythm events, pattern events, beat nodes, and running score. The event association chain verification record is used to compare whether the editing state association relationship and the running state backlink relationship are consistent. The event association chain verification record includes the editing state association identifier and the running state backlink identifier. The editing state association identifier is obtained by combining the rhythm event identifier, pattern event identifier, and beat node identifier formed during the editing stage, and is used to characterize the association relationship between rhythm events, pattern events, and beat nodes before the candidate exported score data is packaged. The running state backlink identifier is obtained by mapping the running events, running beat attribution relationships, and running mode type relationships in the running score through the running backlink result, and is used to characterize the association relationship after the running score is backlinked to the editing state object. When generating the event association chain verification record, the editing state association identifier and the running backlink identifier corresponding to the same running event are written into the same verification record, so that the consistency of the link between rhythm events, pattern events, and beat nodes can be verified item by item in the future.

[0038] The event association chain closure result is obtained by verifying the correspondence between the edit-state association identifier and the runtime backlink identifier. The correspondence verification reads the edit-state association identifier and the runtime backlink identifier from the event association chain verification record, and compares the rhythm event identifier, pattern event identifier, and beat node identifier in the edit-state association identifier with the rhythm event, pattern event, and beat node obtained from the runtime backlink identifier. When the rhythm event identifier, pattern event identifier, and beat node identifier are consistent, the event association chain closure result indicates that the event association chain is closed. When there are inconsistencies in rhythm events, pattern events, or beat nodes, the event association chain closure result indicates that the event association chain is not closed, and the inconsistent identifier positions and corresponding object identifiers are recorded. This verification can identify the situation where the basic backlink has been established but the runtime backlink relationship is inconsistent with the edit-state association relationship, avoiding the situation where the runtime score only shows a successful backlink but the actual association chain is broken.

[0039] Standardized multi-mode spectrogram data or spectrogram link anomaly records are output based on the basic back-connection judgment result and the event association chain closure result. When the basic back-connection judgment result is a pass result and the event association chain closure result indicates that the event association chain is closed, the candidate exported spectrogram data is confirmed as standardized multi-mode spectrogram data. The standardized multi-mode spectrogram data retains the rhythm events, mode events, and the association relationship between mode events and beat nodes, and has passed field integrity verification, internal parsing, back-connection judgment, and event association chain closure verification. When the basic back-connection judgment result is a basic anomaly result, the field integrity anomaly result in the basic back-connection judgment result is read. If the result of the basic reconnection fails, a spectral link anomaly record is generated, containing basic anomaly information, the location of the anomaly field, and the corresponding object identifier. If the basic reconnection judgment result is a basic pass result but the event association chain closure result indicates that the event association chain is not closed, the inconsistency identifier location and the corresponding object identifier in the event association chain closure result are read to generate a spectral link anomaly record containing event association chain anomaly information, the location of the anomaly field, and the corresponding object identifier. The spectral link anomaly record is used to locate the fields and objects in the candidate exported spectral data that cause anomalies in the running parsing, basic reconnection, or event association chain closure, and serves as the basis for determining the spectral objects to be corrected subsequently.

[0040] After verification, candidate exported score data can distinguish closed links that can be used as standardized multi-mode score data output, as well as abnormal links with missing fields, failed backlinks, or unclosed event association chains. Score link abnormality records retain basic abnormal information or event association chain abnormal information, as well as the location of abnormal fields and corresponding object identifiers, providing a basis for subsequent correction and reinforcement learning training. In application: When creating bubble mode, track mode, and beat mode scores for the same target music, candidate exported score data enters internal parsing after field integrity verification, generating a running score containing bubble click running events, track path running events, and beat channel button running events. If all three running events can backlink to the corresponding rhythm event, beat node, and mode event, and the running state backlink identifier is consistent with the editing state association identifier, then standardized multi-mode score data is output. If the beat channel button running event cannot backlink to the corresponding mode event, or the beat node backlinked in the running state is inconsistent with the beat node associated in the editing state, then a score link abnormality record is generated, recording the location of abnormal fields and corresponding object identifiers.

[0041] S7. After generating spectral link anomaly records, generate reinforcement learning training records; refer to the appendix of the instruction manual. Figure 1 and Figure 6The process transforms anomalous objects, correction operations, and corrected verification results in candidate exported spectrogram data into reinforcement learning training records that can be used to update mode event transition strategies. Spectrogram link anomaly records indicate the locations of basic back-connection anomalies or unclosed event association chains. However, simply outputting anomaly records cannot automatically reduce identical spectrogram link anomalies in subsequent similar spectrogram editing data. This implementation first determines the spectrogram objects to be corrected based on the anomaly information and generates reinforcement learning state data by combining it with spectrogram editing correction records. Then, reinforcement learning action data is formed based on mode event transition records and association records. Subsequently, the reinforcement learning action data is applied to the corresponding correction process and the candidate exported spectrogram data is re-verified. Reward data is calculated based on the corrected spectrogram verification results, ultimately forming reinforcement learning training records that include reinforcement learning state data, reinforcement learning action data, and reward data.

[0042] The anomaly records in the spectrogram link determine the spectrogram objects to be corrected, and the enhancement learning state data is generated in conjunction with the spectrogram editing correction records. The spectrogram link anomaly records come from the aforementioned running back-link judgment and event association chain closure verification process. Basic anomaly information indicates anomalies such as field integrity anomalies, running events failing to back-link, running beat attribution relationships failing to back-link, or running mode type relationships failing to back-link. Event association chain anomaly information indicates anomalies such as inconsistencies between the edit state association identifier and the running state back-link identifier. The anomaly field position indicates the position of the field where the anomaly occurred in the candidate exported spectrogram data, running spectrogram, or event association chain verification record. The corresponding object identifier indicates the rhythm event identifier, mode event identifier, or beat node identifier associated with the anomaly field. During processing, the basic anomaly information, event association chain anomaly information, anomaly field position, and corresponding object identifier in the spectrogram link anomaly records are read, and the rhythm event, mode event, or beat node pointed to by the corresponding object identifier is determined as the spectrogram object to be corrected. When the anomaly field position is located in the candidate exported spectrogram data, the corresponding object identifier is used to determine the relevant event, mode event, or beat node. The rhythm events, pattern events, or beat nodes of the link are identified as the chart objects to be corrected, and the position of the abnormal field is taken as the field to be corrected of the chart object. The chart editing correction record is read from the correction operation record formed by the editor for the chart object to be corrected. The chart editing correction record reflects at least the correction object, the correction field, the content before correction, and the content after correction. Then, the chart object to be corrected, the chart link abnormal record, and the chart editing correction record are bound together to form reinforcement learning state data. The reinforcement learning state data is used to represent the abnormal state, abnormal position, abnormal object, and corresponding correction content before the correction occurs, providing the state basis of the same training sample for subsequent reinforcement learning action data and reward data. For example, when the channel button object of the music mode cannot be reconnected due to the absence of the pattern event identifier, the chart object to be corrected is the pattern event corresponding to the channel button object. The chart editing correction record records the operation of adding the pattern event identifier or correcting the pattern type field, and the reinforcement learning state data records the abnormality and the field changes before and after correction.

[0043] Reinforcement learning action data is generated based on pattern event transformation records, association records between pattern events and rhythm events, and score editing correction records. Pattern event transformation records, derived from the aforementioned pattern event transformation process, record the pattern transformation template, object transformation entry point, and transformed pattern event, illustrating the transformation path and field writing source used by the anomalous object during the transformation phase. Association records between pattern events and rhythm events, also derived from the aforementioned association process, record the correspondence between rhythm event identifiers, pattern event identifiers, and beat node identifiers, illustrating the association position of the anomalous object during the relation configuration phase. Score editing correction records provide correction fields and corrected content, determining whether the correction operation should be applied to the pattern event transformation process, relation configuration process, or candidate derived field correction process. During processing, the correction fields in the score editing correction records are compared with the pattern transformation template, object transformation entry point, and pattern event fields in the pattern event transformation records; the correction fields belong to the pattern... When generating event fields, reinforcement learning action data is generated to act on the pattern event transformation process. The correction field is compared with the association records between pattern events and rhythm events. If the correction field belongs to the correspondence between rhythm events, pattern events, or beat nodes, reinforcement learning action data is generated to act on the relationship configuration process. The correction field is compared with the common field area or pattern field area of ​​the candidate exported score data. If the correction field belongs to the exported field content, reinforcement learning action data is generated to act on the candidate exported field correction process. The reinforcement learning action data is used to characterize the position, object, and method of the correction operation in the score production process, providing an action basis for regenerating candidate exported score data. For example, if the event association chain is not closed due to the path termination position not being correctly written to the relationship position marker, the reinforcement learning action data points to the relationship configuration process and records the action of correcting the path termination position to the corresponding beat node range.

[0044] After applying reinforcement learning action data to the corresponding correction process, candidate exported spectrogram data is regenerated, and reward data is calculated based on the corrected spectrogram verification results to form reinforcement learning training records. When reinforcement learning action data is applied to the pattern event transformation process, the fields corresponding to the pattern transformation template and object transformation entry are corrected according to the action data, or the content of the pattern event fields is regenerated, and then the pattern events and candidate exported spectrogram data are regenerated. When reinforcement learning action data is applied to the relation configuration process, the association records between pattern events and rhythm events or the correspondence between pattern events and beat nodes are corrected according to the action data, and then the candidate exported spectrogram data is repackaged. When reinforcement learning action data is applied to the candidate exported field correction process, the corresponding fields in the common field area or pattern field area are corrected according to the action data, and then the candidate exported spectrogram data is regenerated. The regenerated candidate exported spectrogram data continues to perform internal parsing, run backlink judgment, and event association chain closure verification to obtain the corrected spectrogram verification results. The corrected spectrogram verification results include the corrected run backlink results, corrections, and other relevant information. The reward data is generated based on the difference between the pre-correction spectrum link anomaly record and the post-correction spectrum verification result. During calculation, first, the failed back-connection items, unclosed event association chain items, and abnormal field items in the pre-correction spectrum link anomaly record are counted, and then the corresponding abnormal items still existing in the post-correction spectrum verification result are counted. When the running event, running beat attribution relationship, or running mode type relationship changes from being unable to back-connect to being able to back-connect, the reward data is recorded as a back-connection improvement result. When the event association chain closure result changes from unclosed to closed, the reward data is recorded as a link closure improvement result. When the number of abnormal items after correction is less than the number of abnormal items before correction, the reward data is recorded as an anomaly reduction result. When the same anomaly still exists or a new anomaly is added after correction, the reward data is recorded as an unimproved result. The reinforcement learning training record consists of reinforcement learning state data, reinforcement learning action data, and reward data, used for subsequent update mode event conversion strategies to reduce the number of identical spectrum link anomalies in subsequent conversions, relationship configurations, or field encapsulations for similar edited objects.

[0045] In application: When the button objects of the same target music's "Energy Mode" channel fail to reconnect after candidate export, the score link anomaly record indicates that the abnormal field is the mode type field, and the corresponding object identifier points to the mode event of the button object in that channel. After the editor corrects the mode type field, the animation and game production engine software generates the abnormal state and correction record as reinforcement learning state data, and then forms reinforcement learning action data that acts on the mode event conversion process based on the original mode event conversion record. After regenerating the candidate exported score data and completing internal parsing, run reconnection judgment, and event association chain closure verification, if the corrected run reconnection result is valid and no identical score link anomaly record is generated, the reward data is recorded as the improvement result, and together with the reinforcement learning state data and reinforcement learning action data, they form the reinforcement learning training record.

[0046] S8. After generating reinforcement learning training records, update and verify the mode event transition strategy; refer to the appendix of the instruction manual. Figure 1 and Figure 6 The validated anomaly correction experience is reused in the pattern event transformation process of subsequent spectrogram editing data. The reinforcement learning training records already contain anomaly states, correction actions, and reward data, reflecting whether a certain type of spectrogram link anomaly has improved after specific correction actions. Therefore, this implementation process first updates the pattern event transformation strategy based on the reinforcement learning training records, then applies the updated strategy to the pattern event transformation process of subsequent spectrogram editing data. Subsequently, internal parsing, backlink judgment, and event association chain closure verification are performed on the updated candidate exported spectrogram data, and the results are compared with the spectrogram verification results before the update. The updated pattern event transformation strategy is only activated when the backlink status improves or the number of anomaly records decreases after the update. This implementation process includes the following steps: Based on the reinforcement learning training record update pattern event transformation strategy, the updated pattern event transformation strategy is obtained. The reinforcement learning training record comes from the aforementioned reinforcement learning training record generation process, including reinforcement learning state data, reinforcement learning action data, and reward data. The reinforcement learning state data is used to characterize the abnormal object, abnormal field position, and abnormal type when a spectrum link anomaly occurs. The reinforcement learning action data is used to characterize how the correction operation is applied to the pattern event transformation process, relation configuration process, or candidate derived field correction process. The reward data is used to characterize the improvement of the corrected spectrum verification result compared to the original spectrum link anomaly record. The pattern event transformation strategy is used to determine the pattern transformation template, object transformation entry, and field writing rules corresponding to the current editing mode and the type of edited object. During the update process... The system reads records from the reinforcement learning training logs where reward data represents improvement results, and matches the corresponding reinforcement learning action data with the reinforcement learning state data to determine the current editing mode, edit object type, abnormal field location, and correction action that generated the improvement result. When the improvement result corresponds to a pattern event transformation process, the corresponding correction action is written into the pattern transformation template or object transformation entry in the pattern event transformation strategy. When the improvement result corresponds to a relationship configuration process, the corresponding correction action is written into the relationship configuration rules between pattern events and rhythm events. When the improvement result corresponds to a candidate exported field correction process, the corresponding correction action is written into the candidate exported field writing rules. Reinforcement learning action data where reward data represents no improvement results is not written into the preferred transformation rules, thus obtaining the updated pattern event transformation strategy.

[0047] The updated mode event conversion strategy is applied to the mode event conversion process corresponding to subsequent chart editing data, generating updated candidate export chart data. Subsequent chart editing data comes from editing operation records generated via the human-computer interaction interface within the same animation / game production engine software. This data still includes the current editing mode, editing trigger position, editing object type, and editing object relationship data. During application, the updated mode event conversion strategy is first located based on the current editing mode and editing object type in the subsequent chart editing data. Then, mode event conversion is executed according to the updated mode conversion template and object conversion entry point, generating mode events corresponding to the game mode. If the updated mode event conversion strategy... If the updated relational configuration rules are included, the association between pattern events and rhythm events, as well as the association between pattern events and beat nodes, will be established according to the updated relational configuration rules. If the updated pattern event conversion strategy includes candidate exported field writing rules, the common field area and pattern field area will be written according to the updated field writing rules when forming the fieldized score encapsulation record. After completing the pattern event conversion, relational configuration, and field encapsulation, the fieldized score encapsulation record will be bound to the music identifier and editing version identifier corresponding to the target music to generate updated candidate exported score data. The updated candidate exported score data is used to verify the parsing and backlinking effect of the updated pattern event conversion strategy on subsequent score editing data.

[0048] The updated candidate exported chart data undergoes internal parsing, runtime backlink judgment, and event association chain closure verification to obtain the updated chart verification result, which is then compared with the pre-update chart verification result within the same subsequent chart editing data range. Internal parsing reads the updated candidate exported chart data according to the internal parsing order corresponding to the game engine's parsing rules, generating the updated running chart. The runtime backlink judgment maps the running events, running beat attribution relationships, and running mode type relationships in the updated running chart to rhythm events, beat nodes, and mode events, respectively, obtaining the updated running backlink result. The event association chain closure verification obtains the updated event association chain closure result based on the correspondence between the edit state association identifier and the running state backlink identifier. The updated chart verification result includes the updated running backlink result, the updated event association chain closure result, and the updated chart link anomaly record. Record the following: During the comparison process, using the same subsequent score editing data, candidate exported score data are generated according to both the pre-update and post-update mode event conversion strategies. The same internal parsing, backlink judgment, and event association chain closure verification are then performed to obtain the pre-update and post-update score verification results. When the post-update score verification result shows an improvement in the backlink status of running events, running beat attribution relationships, or running mode type relationships compared to the pre-update score verification result, or a reduction in the number of abnormal score link records, the number of abnormal field positions, or the number of related corresponding object identifiers, the updated mode event conversion strategy is activated. When the comparison result does not indicate an improvement in the post-update backlink status or a reduction in abnormal records, the pre-update mode event conversion strategy is retained to prevent unverified strategies from entering the subsequent score production process.

[0049] In application: When creating channel key objects for the same target music in the "Rhythm Mode" subsequent times, if the reinforcement learning training records indicate that correcting the mode type field can resolve the issue of the inability to reconnect the running mode type relationship, the animation and game production engine software will write this correction rule into the updated mode event conversion strategy. When subsequent channel key objects of the same type enter the mode event conversion process, the updated mode event conversion strategy will automatically write the corresponding mode type field and generate updated candidate exported chart data. After internal parsing, running reconnection judgment, and event association chain closure verification, if the updated chart verification results show that the running mode type relationship can reconnect to the mode event and the number of abnormal chart link records is reduced, the updated mode event conversion strategy will be enabled; otherwise, the previous mode event conversion strategy will be retained.

[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for creating music charts for multi-mode music games, characterized in that, include: S1. Obtain the beat data, game mode set, and chart editing data of the target music. The chart editing data includes the current editing mode belonging to the game mode set, the editing trigger position, the type of editing object, and the relationship data of the editing objects. S2. Generate a beat sequence based on the beat data. The beat sequence includes multiple beat nodes arranged in the order of playback of the target music and with event association identifiers. S3. Based on the edit trigger position and the relationship data of the edit object, determine the range of the event in the beat sequence, and generate rhythm events for uniformly associating the same rhythm expression in different game modes. The rhythm events are associated with the beat nodes within the range of the event. S4. Based on the current editing mode and the type of the edited object, convert the score editing data into mode events corresponding to the game mode, and associate the mode events with the corresponding rhythm events. The mode events record the mode type, event performance relationship, and event judgment relationship. S5. Based on the rhythm events and the pattern events associated with the rhythm events, generate preview chart data for the corresponding game modes, and encapsulate the rhythm events, pattern events, and the relationship between the pattern events and the beat nodes into candidate exported chart data. S6. Parse the candidate exported chart data according to the internal parsing rules corresponding to the game engine parsing rules, generate a chart including running events, running beat attribution relationships, and running mode type relationships, map the chart to beat sequences, rhythm events, and mode events, obtain running backlink results, and determine whether running events are backlinked to rhythm events, running beat attribution relationships are backlinked to beat nodes, and running mode type relationships are backlinked to mode events based on the running backlink results, and output standardized multi-mode chart data or chart link anomaly records according to the judgment results.

2. The method according to claim 1, characterized in that, S1 includes: S1.1 Using the music identifier corresponding to the target music as an index, the beat data, game mode set and score editing data are grouped under the same target music to form a dataset to be processed; S1.2 Identify the chart object structure of each game mode in the game mode set and determine the mode structure type corresponding to the current editing mode; S1.

3. According to the mode structure type corresponding to the current editing mode, convert the editing object relationship data in the score editing data into a unified relationship type marker and relationship position marker to obtain the editing record to be processed for the beat sequence.

3. The method according to claim 2, characterized in that, S2 includes: S2.1 Based on the beat speed, beat start time, and beat end time in the beat data, calculate the beat time interval between adjacent beat nodes and the number of beat nodes corresponding to the target music; S2.2 Generate beat nodes corresponding to the number of beat nodes according to the target music playback order, and configure beat node identifier and node time range for each beat node; S2.3 Configure event association identifiers for each beat node to record rhythm event identifiers, and generate beat sequences.

4. The method according to claim 3, characterized in that, S3 includes: S3.1 Match the editing trigger position in the score editing data with the node time range of each beat node in the beat sequence to determine the trigger beat node; S3.

2. Based on the triggering beat node, determine the set of relationship extension nodes in the beat sequence according to the relationship data of the edited object, and determine the triggering beat node and the set of relationship extension nodes as the scope of the event; S3.3 Match the event association identifier of the beat node within the event range according to the event range and the relationship data of the edited object. If a matching event association identifier exists, associate the score editing data with the existing rhythm event. If no matching event association identifier exists, generate a new rhythm event and write the new rhythm event identifier into the beat node within the event range.

5. The method according to claim 4, characterized in that, S4 includes: S4.1 Determine the mode conversion template and object conversion entry point based on the current editing mode and the type of the edited object, and generate a mode event conversion task; S4.

2. According to the mode event conversion task, convert the score editing data into mode events of the corresponding game mode, configure mode event identifiers for mode events, calculate the event performance relationship and event judgment relationship of mode events, and record the mode conversion template, object conversion entry and converted mode events to obtain mode event conversion records. S4.3 Write the pattern event identifier into the corresponding rhythm event, and establish an association between the pattern event and the beat nodes within the event's range to generate an association record between the pattern event and the rhythm event.

6. The method according to claim 5, characterized in that, S5 includes: S5.

1. Based on the rhythm event, the pattern event associated with the rhythm event, and the beat node corresponding to the pattern event, generate the preview chart data for the corresponding game mode according to the current editing mode; S5.2 Write the rhythm event identifier, pattern event identifier, beat node identifier and pattern type into the common field area, and write the event performance relationship and event judgment relationship into the pattern field area to form a fielded spectrum encapsulation record; S5.3 Bind the field-based score sheet encapsulation record with the music identifier and editing version identifier corresponding to the target music to obtain candidate exported score sheet data.

7. The method according to claim 6, characterized in that, S6 includes: S6.1 Perform field integrity verification on candidate exported chart data. If the field integrity verification passes, the candidate exported chart data is parsed according to the internal parsing order corresponding to the game engine parsing rules to generate the running chart. If the field integrity verification fails, an abnormal field integrity result is generated. S6.2 After the field integrity check passes and the running spectrum is generated, the running events, running beat attribution relationships, and running mode type relationships in the running spectrum are mapped to rhythm events, beat nodes, and mode events, respectively, to obtain the running back connection results; S6.3 When the running back connection result is obtained, the back connection status corresponding to the running event, running beat attribution relationship and running mode type relationship is determined according to the running back connection result, and the basic back connection judgment result is generated; when the field integrity abnormal result is generated, the field integrity abnormal result is written into the basic back connection judgment result, and the basic back connection judgment result is used as the basis for judging the output of standardized multi-mode spectral data or spectral link abnormal records.

8. The method according to claim 7, characterized in that, S6 further includes: S6.4 When the basic backlink judgment result is the basic pass result, an event association chain verification record including the edit state association identifier and the run state backlink identifier is generated based on the run backlink result, rhythm event, pattern event, beat node and run score. S6.5 After generating the event association chain verification record, verify the correspondence between the edit-state association identifier and the running-state back-link identifier in the event association chain verification record to obtain the event association chain closure result; S6.6 When the basic back-connection judgment result is a basic pass result and the event association chain closure result indicates that the event association chain is closed, output standardized multi-mode spectral data; when the basic back-connection judgment result is a basic abnormal result, generate a spectral link abnormal record containing basic abnormal information, abnormal field position and corresponding object identifier; when the basic back-connection judgment result is a basic pass result but the event association chain closure result indicates that the event association chain is not closed, generate a spectral link abnormal record containing event association chain abnormal information, abnormal field position and corresponding object identifier.

9. The method according to claim 8, characterized in that, After generating the spectral link anomaly record, it also includes S7: S7.

1. Based on the basic anomaly information, event association chain anomaly information, anomaly field position and corresponding object identifier in the spectrum link anomaly record, determine the spectrum object to be corrected, obtain the spectrum editing and correction record formed for the spectrum object to be corrected, and generate reinforcement learning state data based on the spectrum object to be corrected, the spectrum link anomaly record and the spectrum editing and correction record. S7.2 Generate reinforcement learning action data based on pattern event transition records, correlation records between pattern events and rhythm events, and score editing and correction records; S7.3 Apply the enhanced learning action data to the pattern event transformation process, relation configuration process, or candidate derived field correction process to regenerate candidate derived spectrum data. Perform internal parsing, run back connection judgment, and event association chain closure verification on the regenerated candidate derived spectrum data to obtain the corrected spectrum verification result. The corrected spectrum verification result includes the corrected run back connection result, the corrected event association chain closure result, or the corrected spectrum link abnormal record. The reward data is calculated based on the corrected spectrum verification results, forming a reinforcement learning training record that includes reinforcement learning state data, reinforcement learning action data, and reward data.

10. The method according to claim 9, characterized in that, After generating the reinforcement learning training records, it also includes S8: S8.1 Update the pattern event transition strategy based on the reinforcement learning training records to obtain the updated pattern event transition strategy; S8.2 Apply the updated mode event conversion strategy to the mode event conversion process corresponding to the subsequent spectrum editing data to generate updated candidate exported spectrum data; S8.

3. Perform internal parsing, backlink judgment and event association chain closure verification on the updated candidate exported spectrum data to obtain the updated spectrum verification result. Within the same range of subsequent spectrum editing data, compare the updated spectrum verification result with the pre-update spectrum verification result corresponding to the pre-update mode event conversion strategy. If the comparison result indicates that the updated backlink status is improved or the number of abnormal records is reduced after the update, the updated mode event conversion strategy will be enabled; otherwise, the pre-update mode event conversion strategy will be retained.

Images