A time-slice-based pre-fetching based event playback data scheduling method and system thereof

Abstract

This application discloses a method and system for scheduling sports replay data based on time-sliced ​​prefetching. By dividing the sports replay source data into time slices and performing hierarchical encoding, a mapping relationship between time slices and hierarchical data is constructed, and a basic playback state and hierarchical loading state are established. The prefetch window range is determined based on the target playback position, and the prefetch order is determined by combining time distance and hierarchical prefetching scheduling is executed. A resource allocation model is constructed by combining network bandwidth status, cache occupancy status, and playback rate information to generate scheduling priority parameters and dynamically adjust the hierarchical data loading ratio. Enhanced layer data is delayed and supplemented, and merged and updated with core layer data. Simultaneously, the scheduling process is adjusted in real time based on user playback behavior, enabling the sports replay data to achieve progressive loading of hierarchical data and coordinated optimization of resource utilization while ensuring playback continuity.

Description

Technical Field

[0001] This application relates to the field of video data processing technology, specifically to a method and system for scheduling event replay data based on time-slice prefetching. Background Technology

[0002] With the development of live sports broadcasts and replay services, users have placed higher demands on the real-time performance, continuity, and picture integrity of video replays. Especially in scenarios involving frequent dragging, rewinding, and switching between multiple time periods, ensuring playback continuity while maintaining picture quality under limited network resources has become a key technical challenge in sports replay systems. Against this backdrop, a technical solution is needed that can dynamically schedule data loading during playback, accommodate different levels of data representation, and adapt to varying resource environments to improve the overall replay experience.

[0003] In reality, sports replay data is typically loaded using a linear prefetching method in chronological order. This involves downloading subsequent data in advance using a fixed-size prefetch window and relying on caching mechanisms to maintain playback continuity. Regarding resource allocation, most solutions control data loading based on preset rules or single bandwidth conditions, paying little attention to cache usage and playback rate changes. In terms of data organization, complete video segments are usually used as the basic loading unit, without layered processing of video data. Regarding playback behavior adaptation, existing systems generally re-establish the loading process after a user triggers a drag or jump, lacking a continuous awareness and dynamic response mechanism for user behavior.

[0004] However, the above technical solutions still have shortcomings in practical applications. Specifically, during the loading of event replay data, there are problems such as the lack of time distance constraints in pre-fetching scheduling, the failure to dynamically adjust resource allocation based on bandwidth and cache status, low loading efficiency due to overall video data loading, and the inability to perform real-time scheduling based on user playback behavior. As a result, it is difficult to balance playback continuity and picture integrity in complex network environments and under conditions of changing playback behavior. Summary of the Invention

[0005] In view of the above-mentioned actual situation, this application proposes a method and system for scheduling event replay data based on time-slice prefetching, in order to solve the problems in the prior art that the prefetching scheduling lacks time distance constraints, the resource allocation is not dynamically adjusted in combination with bandwidth and cache status, the overall loading of video data leads to low loading efficiency, and the inability to perform real-time scheduling based on user playback behavior.

[0006] A method for scheduling match replay data based on time-sliced ​​prefetching includes the following steps:

[0007] S1. Obtain the event replay source data, divide the source data along the playback time axis to obtain multiple time slice data arranged in chronological order; perform layered encoding on each time slice to generate core layer data and enhancement layer data, and establish a mapping relationship between time slices, layered data and playback time position; construct a sliced ​​layered data set based on the mapping relationship, and establish a basic playback state to characterize playback continuity and a layered loading state to characterize the loading progress of layered data;

[0008] S2. Based on the target time segment position corresponding to the current playback request, determine the time segments within the prefetch window range, and determine the prefetch order based on the time distance of each time segment relative to the target time segment position; perform hierarchical prefetch scheduling based on the segment-level data set according to the prefetch order, prioritize loading core layer data to establish a continuous playable data sequence, update the basic playback state to a playable state, and mark the enhancement layer data as pending loading.

[0009] S3. Obtain network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Based on the information and the time distance of each time segment relative to the current playback position, determine the resource allocation status corresponding to the time segment and its core layer data and enhancement layer data, and generate scheduling priority parameters. According to the scheduling priority parameters, dynamically adjust the loading ratio of core layer data and enhancement layer data of the time segment within the prefetch window, and update the layer loading status.

[0010] S4. Based on the layer loading status and the scheduling priority parameter, perform delayed supplementary loading of the enhancement layer data to be loaded; during the loading process, gradually load the enhancement layer data according to the time slice order and the scheduling priority parameter, and merge and update it with the already loaded core layer data to improve the picture integrity status of the corresponding time slice.

[0011] S5. During the replay of the event, based on the user's playback behavior information and the basic playback status, hierarchical loading status, and resource allocation status, the prefetch window range, time slice prefetch order, and scheduling priority parameters are adjusted in real time, and the hierarchical prefetch scheduling and delayed supplementary loading are continuously executed to achieve scheduling control of gradually improving the integrity of the replay image during playback.

[0012] Furthermore, the step of determining multiple time segments within the prefetch window range based on the target time segment position corresponding to the current playback request, and determining the prefetch order based on the time distance of each time segment relative to the target time segment position, includes:

[0013] S201. Based on the target time segment position, construct a time distance function space along the playback time axis. Using the target time segment as a reference point, express the time segments on the time axis by distance quantization. In the time distance function space, determine the prefetch window range according to a preset window function, and select multiple time segments within the prefetch window range to form a prefetch candidate set.

[0014] S202. Based on the time distance function space, assign a corresponding time distance weight to each time slice in the prefetch candidate set, and construct a priority sequence of time slices according to the time distance weight; perform sorting processing on the prefetch candidate set according to the priority sequence to generate a prefetch order that satisfies the time continuity constraint, wherein the time distance weight has a monotonic mapping relationship with the time distance of the time slice relative to the target time slice position.

[0015] Furthermore, the acquisition of network bandwidth status information, terminal cache occupancy status information, and current playback rate information, and the determination of resource allocation status and scheduling priority parameters, includes:

[0016] S301. Perform bandwidth time-series modeling processing on the network bandwidth status information to obtain a bandwidth change sequence; perform cache capacity status modeling processing on the terminal cache occupancy status information to obtain a cache occupancy sequence; and construct a data consumption rate model by combining the current playback rate information; based on the bandwidth change sequence, the cache occupancy sequence, and the data consumption rate model, generate a resource status vector to characterize the supply and demand relationship of system resources.

[0017] S302. Based on the resource state vector and the position distribution of each time slice in the time distance function space, construct a multi-dimensional weight allocation model for the core layer data and enhancement layer data of each time slice; according to the multi-dimensional weight allocation model, map system resources to each time slice and its hierarchical data, generate corresponding resource allocation states, and determine the scheduling priority parameters of the hierarchical data based on the resource allocation states, wherein the multi-dimensional weight allocation model includes at least the combination relationship of time distance weight, cache occupancy weight and bandwidth allocation weight.

[0018] Furthermore, the process of performing delayed supplementary loading on the enhancement layer data to be loaded and merging and updating it with the core layer data includes:

[0019] S401. Based on the hierarchical loading state and the scheduling priority parameter, construct a segmented loading sequence for the enhancement layer data in the waiting-to-load state, and assign weights to the segmented loading sequence according to the scheduling priority parameter to form a batch loading strategy for the enhancement layer data. The batch loading strategy is used to control the loading order and loading granularity of the enhancement layer data in the time slice dimension.

[0020] S402. During the data loading process of the enhancement layer, the enhancement layer data is fused with the core layer data of the corresponding time slice, and the data representation of the time slice is iteratively updated based on the fusion result to form a gradually evolving data representation state; the fusion process includes weighted fusion based on hierarchical weights or layered replacement processing based on data coverage relationship, and the image integrity state is synchronously adjusted during the update process.

[0021] Furthermore, the prefetch window range is dynamically adjusted according to the current playback status, including: determining the time span parameter of the prefetch window based on the current playback rate and cache occupancy status, and expanding or shrinking the prefetch window range according to the time span parameter.

[0022] Furthermore, the time distance between the time slice and the current playback position is quantified and calculated using a time distance function. This time distance function is used to establish the priority order relationship of the time slices during the scheduling process and participates in the generation process of the scheduling priority parameters.

[0023] Furthermore, the user playback behavior information includes at least one of drag operations, pause duration, and replay frequency, and the scheduling priority parameter is modified based on the user playback behavior information so that the user playback behavior information participates in the time-slicing scheduling priority calculation process.

[0024] Furthermore, the core layer data and the enhancement layer data are stored in different cache areas, and data access scheduling processing is performed on the cache areas based on the layer loading state, including:

[0025] The core layer data is mapped to the first cache space, the enhancement layer data is mapped to the second cache space, and corresponding access weights are assigned to the first cache space and the second cache space respectively.

[0026] A cache access weight function is constructed based on the hierarchical loading state. The cache access weight function is used to quantify and allocate the access frequency of the first cache space and the second cache space according to the hierarchical loading state, and to constrain the cache read order according to the access weight, so that the core layer data and the enhancement layer data form a hierarchical access order relationship during the cache access process.

[0027] Furthermore, the fusion process includes:

[0028] The core layer data is represented as a first feature vector, and the enhancement layer data is represented as a second feature vector. A hierarchical weight function is constructed in a unified feature space.

[0029] Based on the hierarchical weight function, the corresponding weight coefficients are determined, and the first feature vector and the second feature vector are weighted and combined according to the weight coefficients to obtain the fused feature vector;

[0030] Based on the difference distribution between the fused feature vector and the first feature vector, a data overlay mask is constructed, and a hierarchical replacement operation is performed on the fused feature vector based on the data overlay mask to update the data representation of the corresponding time slice.

[0031] Furthermore, this application also discloses a sports replay data scheduling system based on time-slice prefetching, characterized in that the system includes:

[0032] The acquisition unit is used to acquire the source data of the event replay, divide the source data along the playback time axis to obtain multiple time slice data arranged in chronological order; perform hierarchical encoding on each time slice to generate core layer data and enhancement layer data, and establish a mapping relationship between time slices, hierarchical data and playback time position; construct a slice hierarchical data set based on the mapping relationship, and establish a basic playback state to represent playback continuity and a hierarchical loading state to represent the loading progress of hierarchical data.

[0033] The prefetch scheduling unit is used to determine the time slices within the prefetch window range according to the target time slice position corresponding to the current playback request, and to determine the prefetch order according to the time distance of each time slice relative to the target time slice position; based on the slice-level data set, it performs hierarchical prefetch scheduling according to the prefetch order, prioritizes loading core layer data to establish a continuous playable data sequence, updates the basic playback state to a playable state, and marks the enhancement layer data as pending loading.

[0034] The resource allocation unit is used to acquire network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Based on the information and the time distance of each time segment relative to the current playback position, it determines the resource allocation status of the time segment and its core layer data and enhancement layer data, and generates scheduling priority parameters. According to the scheduling priority parameters, it dynamically adjusts the loading ratio of core layer data and enhancement layer data of the time segment within the prefetch window, and updates the layer loading status.

[0035] The enhancement loading unit is used to perform delayed supplementary loading of the enhancement layer data to be loaded based on the layer loading status and the scheduling priority parameter; during the loading process, the enhancement layer data is loaded step by step according to the time slice order and the scheduling priority parameter, and merged and updated with the already loaded core layer data to improve the picture integrity status of the corresponding time slice.

[0036] The dynamic scheduling unit is used to adjust the prefetch window range, time slice prefetch order, and scheduling priority parameters in real time during the replay of the event, based on user playback behavior information, the basic playback state, hierarchical loading state, and resource allocation state. It also continuously executes the hierarchical prefetch scheduling and delayed supplementary loading to achieve scheduling control that gradually improves the completeness of the replay image during playback.

[0037] The proposed method and system for scheduling event replay data based on time-sliced ​​prefetching realizes the hierarchical data representation and mapping relationship construction based on time slices. It controls the hierarchical data loading process through a time distance-constrained prefetching scheduling mechanism and a multi-dimensional resource allocation model, and dynamically adjusts the scheduling parameters during user playback. This enables the event replay data to achieve progressive loading of hierarchical data and coordinated optimization of resource utilization while ensuring playback continuity. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating the event replay data scheduling method based on time-slicing prefetching proposed in this application.

[0039] Figure 2 This is a schematic diagram of the time-slicing and hierarchical mapping structure in this embodiment;

[0040] Figure 3 This is a schematic diagram illustrating the relationship between the prefetch window and the target playback position in this embodiment;

[0041] Figure 4 This is a schematic diagram illustrating the progressive loading process of the core layer data and the enhancement layer data in this embodiment;

[0042] Figure 5 A schematic diagram of the logical structure for generating the resource state vector and scheduling priority parameters in this embodiment;

[0043] Figure 6 A schematic diagram of a sports replay data scheduling system based on time-slicing prefetching is provided in an embodiment of this application; Detailed Implementation

[0044] The simulation technology route in 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.

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0046] The features and performance of the present invention will be further described in detail below with reference to embodiments. Please refer to the appendix. Figure 1 The above describes a method for scheduling match replay data based on time-sliced ​​prefetching, which includes the following steps:

[0047] S1. Obtain the event replay source data, divide the source data along the playback time axis to obtain multiple time slice data arranged in chronological order; perform layered encoding on each time slice to generate core layer data and enhancement layer data, and establish a mapping relationship between time slices, layered data and playback time position; construct a sliced ​​layered data set based on the mapping relationship, and establish a basic playback state to characterize playback continuity and a layered loading state to characterize the loading progress of layered data;

[0048] In some implementations, this step involves establishing a basic data organization structure around the event replay source data for subsequent hierarchical prefetching scheduling. Specifically, the event replay source data is denoted as V, and it is expanded along the playback time axis into a continuous sequence of video content and its corresponding time index sequence. The time index sequence is used to characterize the positional relationship of each video content unit during playback. Furthermore, to ensure that subsequent prefetch window determination, prefetch order calculation, and resource allocation calculation have a unified data object, the event replay source data needs to be divided according to a preset time span, forming multiple time slices arranged in chronological order. It should be noted that the time slices are not isolated extractions of the source data, but rather each time slice corresponds to a defined playback time interval, thus giving each time slice both data carrying attributes and playback position attributes. Based on this, the total duration of the event replay source data is denoted as T, and the length of each time slice is... The total number of time slices N is expressed as Where T represents the total playback duration of the event replay source data. This indicates the time span corresponding to a single time slice. This represents the floor function; therefore, the i-th time slice... The corresponding playback time range is represented as ,in This represents a continuous time variable on the playback timeline. This indicates that the event replay source data is in time. The data content at the location; as for the alignment processing of frame-level content, timestamp correction processing and fragment container encapsulation processing of time fragment boundaries, the calculation or processing methods are well known to those skilled in the art and will not be elaborated on here.

[0049] In this embodiment, after time-slicing is completed, each time slice undergoes layered encoding to generate core layer data and enhancement layer data. The core layer data carries low-level data components that constitute the basic playable image, while the enhancement layer data carries data components that supplement image texture, motion details, and local change information. That is, the core layer data corresponds to the data set that can independently support the basic image output when bandwidth is limited or the loading process is not yet complete, and the enhancement layer data corresponds to the data set that participates in detail restoration when performing layer-by-layer supplementation on top of the core layer data. Specifically, for the i-th time slice... First, through layered coding operators The decomposition relationship between the basic representation and the detailed representation is denoted as: ,in This represents the core layer data corresponding to the i-th time slice. This represents the enhancement layer data corresponding to the i-th time slice; further, the layered coding process can adopt a scalable video coding method, a residual layered construction method, or a two-layer data generation method based on feature decomposition. In this embodiment, the layered coding process is expressed in the manner of "basic reconstruction component - detail residual component", that is, in the original slice representation... Representation of basic reconstruction Establish residual relationships between them ,in This represents the original data representation of the i-th time slice. This represents the basic image representation reconstructed based on the core layer data. This represents the detail residual between the original data representation and the basic image representation; the core layer data The enhancement layer data is composed of the encoding results corresponding to the basic image representation. The enhancement layer data is composed of the encoding results corresponding to the detailed residuals, thereby making the enhancement layer data logically subordinate to the core layer data and providing a separable data source for subsequent fusion and update processing. The process of using a specific encoder to implement the basic image representation and detailed residual encoding is a well-known implementation method for those skilled in the art, and will not be elaborated on here.

[0050] In some implementations, the division between the core layer data and the enhancement layer data does not merely correspond to a split in data volume, but rather to two different types of objects acting in playback modeling. The core layer data is used to establish the basic playback state, while the enhancement layer data is used to drive the evolution of the hierarchical loading state. Therefore, after layered encoding is completed, it is also necessary to establish a mapping relationship between time slices, hierarchical data, and playback time positions to form a unified index structure called in subsequent steps. Further, let the playback time position of the i-th time slice be denoted as... ,but ,in This represents the starting position of the i-th time slice on the playback timeline; based on this, a mapping function is constructed. This binds the time segment identifier, core layer data, enhancement layer data, and playback time position within a unified relational space, represented as follows: ,in Let represent the mapping term corresponding to the i-th time slice; further, based on all mapping terms, construct a slice-level data set D, denoted as . The sharded hierarchical data set D is used as the basic data source for determining the prefetch window and arranging the prefetch order in the subsequent S2, and also as the data retrieval object for generating resource allocation status and scheduling priority parameters in the subsequent S3, so that the steps are continuously connected along a unified sharded index logic.

[0051] In this embodiment, after establishing the segmented hierarchical data set, it is necessary to further establish a basic playback state to characterize playback continuity and a hierarchical loading state to characterize the loading progress of layered data. The basic playback state describes whether each time segment has met the basic playback conditions, and the hierarchical loading state describes the loading occupancy relationship of each time segment at the core layer data and enhancement layer data levels. Specifically, for the i-th time segment, a basic playback state is defined. When core layer data Once loading is complete and the video is ready for basic decoding, remember... When core layer data Before the conditions for playback are met, remember... Then it can be expressed as ,in This represents the basic playback state of the i-th time slice; further, to ensure that the delayed supplementary loading of subsequent enhancement layer data has quantifiable state entries, the hierarchical loading state of the i-th time slice is defined. for ,in This represents the amount of data loaded into the core layer of the i-th time segment. This represents the loading occupancy of the enhancement layer data in the i-th time slice; further, to unify the hierarchical loading expression between different time slices, the loading occupancy can be normalized into an interval variable, denoted as the total amount of core layer data. The total amount of data in the enhancement layer is Then there is , ,in This represents the amount of core layer data that has been loaded in the i-th time slice. This represents the amount of enhancement layer data already loaded in the i-th time slice; thus, the basic playback state... With the aforementioned hierarchical loading state Together, they constitute the basic state set in subsequent scheduling calculations. The basic state set not only inherits the data organization structure formed in the current step, but also provides state input for hierarchical prefetching scheduling, resource allocation, and enhancement layer supplementary loading in subsequent steps.

[0052] In some implementations, to ensure that the basic playback state and the hierarchical loading state form a unified management object within the system, a time-slice state set Q is further constructed, denoted as... Here, Q represents a state set based on a time-sharded index. The state set Q and the aforementioned sharded hierarchical data set D together constitute the data entry point for subsequent steps. The sharded hierarchical data set D is for sharded retrieval and hierarchical data retrieval, while the state set Q is for scheduling calculation and loading progress updates. The two establish a corresponding relationship at the time-sharded index i. It should be noted that the persistence of the state set, cache registration, and index table maintenance process involves conventional data structure organization methods. The calculation or processing methods are well known to those skilled in the art, and will not be elaborated on here.

[0053] In this embodiment, through the above processing, the segmented hierarchical data set, playback basic state, and hierarchical loading state formed in S1 have clear time position attributes, hierarchical attributes, and state attributes. Subsequently, when S2 determines the prefetch window range and arranges the prefetch order based on the time segment index and playback time position, it can directly call the time position item in the segmented hierarchical data set D. With hierarchical data items , And combined with the basic playback states in the state set Q. With hierarchical loading state Prioritize loading core layer data and handle the loading of enhancement layer data, thereby ensuring a continuous connection between the current step and subsequent steps in terms of both data structure and state structure.

[0054] Please see the appendix Figure 2 As shown, attached Figure 2The diagram illustrates the structure of time-slicing and hierarchical mapping. In this embodiment, the event replay source data is divided into multiple time slices along the playback timeline. These time slices are arranged sequentially on the timeline and mapped to their corresponding playback time positions. Within each time slice, the data is further divided into core layer data and enhancement layer data. The core layer data forms the basic playback data unit, while the enhancement layer data provides detailed supplementation on top of the basic playback data unit. Simultaneously, a state structure representing the hierarchical loading progress is set on the side of each time slice to reflect the loading and occupancy status of the core layer data and enhancement layer data, thereby forming a slice-level hierarchical data set, providing a data foundation for subsequent prefetching scheduling and resource allocation.

[0055] S2. Based on the target time segment position corresponding to the current playback request, determine the time segments within the prefetch window range, and determine the prefetch order based on the time distance of each time segment relative to the target time segment position; perform hierarchical prefetch scheduling based on the segment-level data set according to the prefetch order, prioritize loading core layer data to establish a continuous playable data sequence, update the basic playback state to a playable state, and mark the enhancement layer data as pending loading.

[0056] Furthermore, the step of determining multiple time segments within the prefetch window range based on the target time segment position corresponding to the current playback request, and determining the prefetch order based on the time distance of each time segment relative to the target time segment position, includes:

[0057] S201. Based on the target time segment position, construct a time distance function space along the playback time axis. Using the target time segment as a reference point, express the time segments on the time axis by distance quantization. In the time distance function space, determine the prefetch window range according to a preset window function, and select multiple time segments within the prefetch window range to form a prefetch candidate set.

[0058] S202. Based on the time distance function space, assign a corresponding time distance weight to each time slice in the prefetch candidate set, and construct a priority sequence of time slices according to the time distance weight; perform sorting processing on the prefetch candidate set according to the priority sequence to generate a prefetch order that satisfies the time continuity constraint, wherein the time distance weight has a monotonic mapping relationship with the time distance of the time slice relative to the target time slice position.

[0059] Furthermore, S2 expands upon the segmented hierarchical data set and state set formed in S1. The segmented hierarchical data set records the correspondence between time segments, core layer data, enhancement layer data, and playback time positions. The state set records the basic playback state and hierarchical loading state corresponding to each time segment. Accordingly, the processing object of S2 is no longer the original event replay source data, but a time segment index structure with time position attributes, hierarchical data attributes, and loading state attributes. Specifically, S2 forms a prefetch window on the playback time axis for the target time segment position indicated by the current playback request, and constructs a prefetch order within the prefetch window according to the time distance between the target time segment position and the target time segment position. This provides an ordered entry point for the priority loading of subsequent core layer data, and at the same time provides a time position dimension constraint basis for the calculation of resource allocation state and scheduling priority parameters in subsequent S3.

[0060] In this embodiment, the target time slice position corresponding to the current playback request is denoted as g, where g is the target index value in the time slice index space and corresponds to the mapping function in S1. The corresponding time segment numbers remain consistent; furthermore, the playback time position corresponding to the target time segment position g is denoted as... Therefore, in S2, the determination of the prefetch window does not directly use a fixed number of slices to truncate forward and backward. Instead, a time distance function space is constructed on the time axis around the target time slice position, so that the prefetch window determination process and the prefetch order generation process share a unified quantitative basis. It should be noted that the time distance function space is used to characterize the distance relationship between any time slice and the target time slice position. This distance relationship participates in both the screening of the prefetch candidate set and the sorting calculation of the prefetch order. Therefore, the time distance function constitutes a unified scheduling variable in this step.

[0061] In some implementations, the processing in S201 involves establishing a time distance function space along the playback time axis based on the target time slice position g, and determining the prefetch window range within this time distance function space in conjunction with a window function; specifically, for any time slice... The time distance relative to the target time slice position g is defined as... ,in This represents the time distance between the i-th time slice and the target time slice. Represents the time distance function. This indicates the playback time position of the i-th time segment. This indicates the playback time position corresponding to the target time segment position; it has been defined in S1. Therefore, the above formula can be further expressed as ,in This represents the time span corresponding to a single time slice. Therefore, all time slices are mapped to a set of non-negative distance values ​​relative to the target time slice position g. A one-to-one correspondence exists between this set of non-negative distance values ​​and the time slice index set, thus forming the time distance function space. Furthermore, this time distance function space is not only used to express absolute distance values ​​but also to provide the domain for subsequent window functions and weight functions; therefore, it is essentially a scheduling coordinate space constructed around the target time slice position.

[0062] In this embodiment, to ensure that the process of determining the prefetch window has a clear computable boundary, a window function is constructed within the time distance function space. The time slices are then selected to enter the prefetch candidate set based on this selection. Specifically, given a window radius parameter R, where R represents the maximum time distance threshold relative to the target time slice position that is allowed to be included in the prefetch candidate range, the window function is expressed as follows: ,in The value of is used to characterize whether the i-th time slice falls within the prefetch window range; based on this, the prefetch candidate set P is denoted as Where P represents the prefetch candidate set obtained by window function filtering within the time distance function space; it should be noted that the window radius parameter R can be set by a fixed time span, a time span corresponding to a fixed number of segments, or a playback context-defined interval. The method of selecting its value is well known to those skilled in the art, and will not be elaborated further here. The introduction of the window function makes it no longer dependent on discrete experience to determine whether a time segment enters the prefetch set, but transforms it into a threshold filtering result based on the time distance function space, thereby giving S201 a definite boundary expression in its computational structure.

[0063] In some implementations, the prefetch candidate set P in S201 also needs to satisfy continuity in the time interval to avoid forming discrete, skip-type prefetch objects near the target time slice; specifically, in the window function After filtering, a continuity check can be performed on the time slice indices falling within the window range, and abnormal slice indices that do not meet the requirement of continuous time intervals can be removed, thereby maintaining the continuous interval attribute of the prefetch candidate set on the playback timeline. The index traversal, boundary alignment, and abnormal index correction involved in the continuity check are calculations or processing methods well known to those skilled in the art, and will not be elaborated upon here. Through the above processing, S201 outputs a prefetch candidate set P that is within the prefetch window range and has the attribute of continuous time intervals. Each time slice in this set has a clearly defined time distance value. This provides input for the subsequent assignment of time distance weights and construction of priority sequences in S202.

[0064] In this embodiment, step S202 takes the time distance function space formed in S201 and the prefetch candidate set P as input, assigns a corresponding time distance weight to each time slice in the prefetch candidate set, and constructs a priority sequence based on the time distance weight. Specifically, the time distance weight is not simply the time distance value directly used for sorting, but rather it is determined through a monotonic mapping function. The distance is converted into a weight value that can participate in the priority calculation. Let the time distance weight corresponding to the i-th time slice be denoted as . Then there is ,in The time distance weight function represents a time distance weight that has a monotonic mapping relationship with the time distance. Further, in this embodiment, the monotonic mapping relationship can be implemented using a reciprocal mapping, an exponential decay mapping, or a piecewise decay mapping. If a reciprocal mapping is used, it can be written as... ,in Indicates the distance attenuation coefficient. If an exponentially decaying mapping is used, it can be written as: ,in Represents the exponential decay coefficient. Both of the above forms satisfy the monotonic mapping condition, meaning that when the time distance changes, the time distance weight changes in an ordered manner according to a predetermined functional relationship. This allows time slices that are closer to the target time slice position to obtain higher ranking participation values ​​in the weight space. Regarding the selection of the specific form of the weight function and the parameter tuning method, this calculation or processing method is well-known to those skilled in the art, and will not be elaborated upon here.

[0065] In some implementations, only time distance weighting is used. The sorting process still needs to be combined with time continuity constraints to avoid the sorting results creating a data retrieval order that is detrimental to continuous playback before and after the target time slices; therefore, when constructing the priority sequence, a priority evaluation metric is established for the time slices in the prefetch candidate set. The priority evaluation metric is determined by both the time distance weight and the time continuity constraint, and is expressed as follows: ,in Indicates the priority constructor. This represents the temporal continuity constraint term for the i-th time slice; in this embodiment, the temporal continuity constraint term... This constraint is used to determine how the sorting results expand relative to the target time segment position, ensuring that the sorting process continues to expand continuously towards both sides of the playback time axis around the target time segment position without introducing jumps across intervals. Specifically, the time continuity constraint can be constructed through the difference between adjacent segment indices, continuous interval maintenance identifiers, or local interval connectivity relationships. Its calculation method is well-known to those skilled in the art and will not be elaborated upon here. The priority evaluation quantity is introduced... In S202, the sorting object is transformed from a single distance value into a composite quantity with continuous constraints, so that the prefetching order reflects the sequential relationship within the neighborhood of the target time slice and maintains the data continuity in the continuous playback scenario.

[0066] In this embodiment, the priority evaluation is based on the priority evaluation value corresponding to all prefetched candidate fragments. Construct a priority sequence for time slices. , represented as ,in The sorting operation is performed based on the priority evaluation value; further, the priority sequence Used to determine the time-slice access order during the prefetch scheduling process, that is, the prefetch process sequentially calls the corresponding core layer data in the slice-level data set D along the priority sequence. With enhancement layer data It should be noted that the stable sorting, parallel priority resolution, and boundary index correction processes involved in the sorting operation are implementation methods well known to those skilled in the art, and will not be elaborated upon here. Through the above processing, S202 generates not a simple index arrangement, but a hierarchical prefetch entry sequence that satisfies the time distance constraint and the time continuity constraint.

[0067] In some implementations, after obtaining the prefetch order Subsequently, S2 further performs hierarchical prefetch scheduling based on the fragmented hierarchical data set D according to the prefetch order. In this step, the hierarchical prefetch scheduling prioritizes core layer data. Priority loading is the main processing thread, that is, loading according to the priority sequence. The time slices are arranged in order, and the core layer data is used first. The data is entered into the playback buffer to form a continuous, playable sequence; specifically, if we consider the prefetch order... The k-th time slice in the data is ,in This represents the k-th time shard index after sorting, and the core layer prefetch set. Represented as ,in This indicates the number of time slices in the priority sequence; during this prefetching process, once a certain time slice... Core layer data Once loaded to meet the basic playback conditions, the corresponding playback basic state for that time segment... Updated to Correspondingly, its enhancement layer data In the current step, it is marked as being in a pending loading state, which serves as the input condition for delayed supplementary loading in subsequent S4; further, if the pending loading flag variable is denoted as... For time slices where the core layer has been loaded but the enhancement layer has not yet been loaded, then... ,in This indicates that the enhancement layer data of the i-th time slice has entered the loading set. This indicates that the enhancement layer data for the i-th time slice has not been included in the loading set or has already been processed. Therefore, after S2 completes the core layer priority loading, it adjusts the playback basic state. The process transitions from an unsatisfied state to a playable state, while simultaneously creating an enhancement layer loading flag variable. This provides continuous input at the state level for subsequent resource allocation calculations in S3 and delayed supplementary loading of the enhancement layer in S4.

[0068] In this embodiment, S2, through the progressive processing of S201 and S202, completes the construction of the time distance function space starting from the target time slice position, the determination of the prefetch window, the generation of the prefetch order, and the core layer priority loading processing. The time distance function... It provides a unified foundation for distance quantization and window functions. Provides boundary filtering conditions for the prefetched candidate set and a time distance weighting function. With priority constructor This provides the basis for calculating the prefetch order; therefore, the output of this step should at least include the updated set of basic playback states. Enhancement layer unloaded identifier set and the prefetching order that satisfies the time continuity constraint. Subsequently, when S3 determines the resource allocation status and scheduling priority parameters, it will use the aforementioned prefetch order. The time segmentation arrangement relationship is used as the time position input, and combined with network bandwidth status information, terminal cache occupancy status information and current playback rate information to construct a multi-dimensional resource allocation relationship, thereby realizing the continuous connection between this step and subsequent steps in terms of data structure, sorting structure and state structure. The time distance function, weight function and priority function can be implemented in different mathematical forms, but all satisfy the monotonic mapping relationship constraint.

[0069] Please see the appendix Figure 3 As shown, attached Figure 3The relationship between the prefetch window and the target playback position is illustrated. In this embodiment, a prefetch window range is constructed on the playback timeline, centered on the target time slice position corresponding to the current playback request. The prefetch window range covers multiple time slices within a certain time span before and after the target time slice position. Within this prefetch window range, each time slice has a different time distance relative to the target time slice position, and a corresponding weight distribution relationship is formed based on the time distance. Furthermore, the time slices in the prefetch candidate set are sorted according to the time distance weights to form a prefetch order around the target time slice position, so that time slices that are temporally adjacent participate in data prefetching first, thereby ensuring playback continuity.

[0070] S3. Obtain network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Based on the information and the time distance of each time segment relative to the current playback position, determine the resource allocation status corresponding to the time segment and its core layer data and enhancement layer data, and generate scheduling priority parameters. According to the scheduling priority parameters, dynamically adjust the loading ratio of core layer data and enhancement layer data of the time segment within the prefetch window, and update the layer loading status.

[0071] Furthermore, the acquisition of network bandwidth status information, terminal cache occupancy status information, and current playback rate information, and the determination of resource allocation status and scheduling priority parameters, includes:

[0072] S301. Perform bandwidth time-series modeling processing on the network bandwidth status information to obtain a bandwidth change sequence; perform cache capacity status modeling processing on the terminal cache occupancy status information to obtain a cache occupancy sequence; and construct a data consumption rate model by combining the current playback rate information; based on the bandwidth change sequence, the cache occupancy sequence, and the data consumption rate model, generate a resource status vector to characterize the supply and demand relationship of system resources.

[0073] S302. Based on the resource state vector and the position distribution of each time slice in the time distance function space, construct a multi-dimensional weight allocation model for the core layer data and enhancement layer data of each time slice; according to the multi-dimensional weight allocation model, map system resources to each time slice and its hierarchical data, generate corresponding resource allocation states, and determine the scheduling priority parameters of the hierarchical data based on the resource allocation states, wherein the multi-dimensional weight allocation model includes at least the combination relationship of time distance weight, cache occupancy weight and bandwidth allocation weight.

[0074] Furthermore, S3 expands upon the prefetch order, basic playback state set, hierarchical loading state set, and enhancement layer pending loading identifier set output by S2. The prefetch order represents the temporal scheduling order of each time slice within the prefetch window; the basic playback state set represents whether each time slice has met the basic playback conditions; the hierarchical loading state set represents the loading progress of each time slice at both the core layer data and enhancement layer data levels; and the enhancement layer pending loading identifier set represents which time slices have entered the candidate range for subsequent enhancement layer supplementary loading. Specifically, S3 does not perform isolated calculations on network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Instead, it integrates this information with the temporal order structure, playback state structure, and loading state structure already formed in the preceding steps into a unified resource supply and demand description framework. This determines the resource allocation status corresponding to each time slice and its hierarchical data within the prefetch window and generates scheduling priority parameters to support the delayed supplementary loading process of enhancement layer data in subsequent S4.

[0075] In this embodiment, S301 establishes a unified resource state representation for three types of resource constraint sources: network side, cache side, and playback side. Specifically, bandwidth temporal modeling is performed on the network bandwidth state information, cache capacity state modeling is performed on the terminal cache occupancy state information, and a data consumption rate model is constructed by combining the current playback rate information. The bandwidth temporal modeling is used to characterize the supply trend of network resources within the corresponding time period of the prefetch window, the cache capacity state modeling is used to characterize the spatial constraints on the terminal side's ability to carry prefetched data, and the data consumption rate model is used to characterize the intensity of data consumption by the player at the current playback rate. Further, if the bandwidth sampling sequence is denoted as... The cache occupancy sampling sequence is denoted as The playback speed is recorded as Then a resource state vector can be established around the current scheduling moment. , represented as ,in This represents the bandwidth within the current scheduling interval. This represents the cache usage within the current scheduling interval. Indicates the current playback speed The data consumption rate is a metric that represents the rate of data consumption. This represents the transpose operation; the resource state vector This is used to transform scattered resource supply and consumption information into a unified vector representation. The subsequent multi-dimensional weight allocation model in S302 uses the resource state vector as the resource dimension input.

[0076] In some implementations, the bandwidth representation quantity Cache usage metrics and data consumption rate characterization All calculations are performed using a local time interval adjacent to the current scheduling time to maintain consistency with the time context corresponding to the prefetch window in S2. Specifically, the bandwidth representation... The buffer occupancy can be represented by the mean, weighted mean, or smoothed estimate of the bandwidth variation sequence over a local time interval. The data consumption rate can be obtained by transforming at least one of the cache occupied capacity, remaining capacity, and writable capacity. With the current playback speed and the current playback basic state set The corresponding playable segment density is relevant; it should be noted that bandwidth smoothing, cache capacity conversion, and rate statistics are all routine calculations in player scheduling, and these calculations or processing methods are well-known to those skilled in the art, so they will not be elaborated on here. The focus of processing in step S301 is not on measuring a single resource in isolation, but on establishing a unified entry point for resource supply and demand relationships using the resource state vector, so that subsequent resource mapping is no longer a discrete judgment, but a structured variable that can participate in weight allocation.

[0077] In this embodiment, to further couple the resource supply and demand relationship with the positional distribution of time slices within the prefetch window, S302 introduces the time distance function space already established in S2 based on the resource state vector, and constructs a multi-dimensional weight allocation model for each time slice within the prefetch window. Specifically, each time slice within the prefetch window has already obtained its time distance value in S2. Time distance weight and prefetching order Therefore, S302 no longer regenerates the time-location relationship, but instead uses the time-distance weight. The location dimension is considered as input, and a joint weight is formed by combining the cache dimension component and the bandwidth dimension component in the resource state vector; furthermore, for the i-th time slice, a comprehensive allocation weight can be constructed. , represented as ,in This represents the time distance weight derived from the time distance function space. This represents the cache usage weight derived from the cache usage status. This represents the bandwidth allocation weight derived from the bandwidth supply status. The weight coefficients of the corresponding dimensions are represented and satisfy the non-negativity and normalization constraints. Therefore, the multi-dimensional weight allocation model is not a simple linear sorting, but rather couples the time position dimension, cache capacity dimension, and network supply dimension into the same allocation calculation process, so that each time slice obtains a comprehensive allocation amount corresponding to its position state and resource state.

[0078] In some implementations, the cache occupancy weight Construction and hierarchical loading state collection Construction and hierarchical loading state collection It has been defined in S1 Furthermore, regarding the basic playback state... And the enhancement layer is yet to be loaded. The time-slicing cache usage weight no longer points to the core layer data, but instead shifts to the subsequent supplementary loading requirements of the enhancement layer data, for the basic playback state. The time slices are assigned cache usage weights that are preferentially associated with the loading needs of the core layer data. This results in different cache resource participation relationships for the same time slice under different playback and loading states. It should be noted that the cache usage weights... The specific conversion can be achieved through the proportional relationship, mapping relationship, or normalization relationship between the remaining cache capacity and the data volume of the fragment level. This calculation or processing method is well-known to those skilled in the art, and will not be elaborated upon here. By using the aforementioned basic playback state set... Hierarchical loading state set and the set of identifiers to be loaded in the enhancement layer In the process of constructing cache weights, S302 logically transitions from "time position priority" to "time position-hierarchical state joint priority".

[0079] In this embodiment, the bandwidth allocation weight With prefetch order and the basic playback state set The coupling method is reflected in the segmented mapping of available bandwidth. Specifically, the prefetch order generated by S2... The loading order of time segments within the prefetch window has already been characterized. Bandwidth allocation in S3 is no longer performed independently of this order. Instead, it allocates the current bandwidth supply sequentially to the hierarchical data of the corresponding time segments based on the prefetch order. For the basic playback state... In time-slicing, the bandwidth allocation weight is preferentially applied to the loading share of core layer data; for playback base state And the enhancement layer is yet to be loaded. The bandwidth allocation weights are used in the calculation of the supplementary loading share of enhancement layer data during time slicing. Therefore, the bandwidth allocation weights... This not only reflects the network supply capacity, but also the positional constraints in the prefetching order and the differences in hierarchical requirements in the playback state, so that bandwidth resources are allocated in an orderly manner towards the core layer and the enhancement layer at the fragmentation level.

[0080] In some implementations, the overall allocation weights for each time slice are obtained. Next, the system resources need to be further mapped to each time slice and its hierarchical data to generate the corresponding resource allocation status. Specifically, if the total amount of data resources available for pre-fetching within the current scheduling period is denoted as... The amount of fragment resources allocated to the i-th time fragment is... It can be represented as ,in This represents the total amount of data resources available for prefetching within the current scheduling period, where P represents the prefetch candidate set generated in S201; further, the amount of fragmented resource allocation... It is also necessary to perform a hierarchical split between the core layer data and the enhancement layer data. This splitting process is related to the basic playback state. Hierarchical loading status and reinforcement layer loading identifier The corresponding state conditions are relevant, that is, for Time slices, The data loading component is prioritized for allocation to the core layer. and Time slices, The data loading components of the enhancement layer are allocated according to the loading degree of the enhancement layer; the specific ratio conversion, boundary control and residual amount reclamation of the fragmentation resource allocation between the core layer and the enhancement layer are well known to those skilled in the art and will not be elaborated here.

[0081] In this embodiment, the resource allocation status is used not only to express the resource share obtained by each time slice in the current scheduling period, but also to generate the scheduling priority parameters for hierarchical data. Specifically, for the i-th time slice, its scheduling priority parameter... Based on the joint constraints of the allocated resources for each segment, the basic playback state, and the pending loading flags of the enhancement layer, it is represented as: ,in This indicates the priority parameter generation function, the function This is used to map resource allocation status and playback status together into control variables that participate in adjusting the loading ratio of subsequent layers; furthermore, when At that time, the scheduling priority parameter Participate in the calculation of increasing the core layer data loading ratio, when and At that time, the scheduling priority parameter It participates in the allocation calculation of the enhancement layer data loading ratio, thereby enabling the same priority parameter generation structure to serve both the basic playable establishment process of the core layer data and the subsequent progressive loading process of the enhancement layer data. (Regarding the priority parameter function...) The specific mapping form, normalization form, and boundary truncation form are all well-known to those skilled in the art, and will not be elaborated upon here.

[0082] In some implementations, S3 generates scheduling priority parameters. Subsequently, the loading ratio of core layer data and enhancement layer data in the time slices within the prefetch window needs to be dynamically adjusted according to the scheduling priority parameter, and the hierarchical loading status set needs to be updated synchronously. Specifically, the dynamic adjustment does not re-determine the prefetching order already completed in S2. Instead, it maintains the prefetch order. Under the premise of no change, the hierarchical loading shares of core layer data and enhancement layer data are determined for each time slice within the prefetch candidate set P, thereby transforming the resource allocation state into an executable hierarchical loading quantity; when the core layer data loading quantity of a certain time slice increases, its hierarchical loading state... In Updated proportionally, when the data loading volume of the enhancement layer of a certain time segment increases, its hierarchical loading status changes. Update according to the corresponding ratio; if Once the core layer reaches full load, the basic playback state is... If it remains playable, If the conditions for subsequent enhancement layer supplementary processing are met, then the enhancement layer to be loaded is identified. In subsequent steps, this serves as the entry condition for the delayed supplementary loading of the enhancement layer. Thus, S3 completes the state progression from "quantification of resource supply and demand" to "adjustment of hierarchical loading ratio," and establishes a connection in the resource allocation dimension between the prefetching sequence structure formed in S2 and the hierarchical loading state structure to be invoked in S4.

[0083] In this embodiment, after continuous processing in S301 and S302, the output of S3 includes at least a resource state vector, the comprehensive allocation weight of each time slice, the resource allocation amount of each slice, the scheduling priority parameter, and the updated hierarchical loading state set. The resource state vector addresses how to uniformly express the supply and demand of system resources; the multi-dimensional weight allocation model addresses how time location factors, caching factors, and bandwidth factors jointly participate in the allocation; and the resource allocation state and scheduling priority parameter address how to form differentiated loading ratios for each time slice and its hierarchical data within the prefetch window. Furthermore, in the subsequent S4, when performing delayed supplementary loading of the enhancement layer data to be loaded, the aforementioned scheduling priority parameter is used as the control input for the batch loading order and loading share of the enhancement layer, and the aforementioned updated hierarchical loading state set is used as the status input for the enhancement layer loading progress. This ensures that this step and subsequent steps are continuously connected in three dimensions: time sequence structure, resource allocation structure, and hierarchical state structure.

[0084] Please see the appendix Figure 5 As shown, attached Figure 5 The logical structure for generating resource state vectors and scheduling priority parameters is illustrated. In this embodiment, network bandwidth status information, terminal cache occupancy status information, and current playback rate information are used as inputs. A resource state vector is constructed through aggregation processing to characterize the resource supply and demand status of the system at the current moment. Based on the resource state vector, and combined with the distribution position of time slices in the time distance function space, weight relationships are constructed for the time distance dimension, cache occupancy dimension, and bandwidth allocation dimension, respectively. The system resources are mapped and allocated through a multi-dimensional weight allocation model, thereby generating the resource allocation status and scheduling priority parameters corresponding to each time slice and its hierarchical data. The scheduling priority parameters are used to constrain the loading order and loading ratio of subsequent hierarchical data.

[0085] S4. Based on the layer loading status and the scheduling priority parameter, perform delayed supplementary loading of the enhancement layer data to be loaded; during the loading process, gradually load the enhancement layer data according to the time slice order and the scheduling priority parameter, and merge and update it with the already loaded core layer data to improve the picture integrity status of the corresponding time slice.

[0086] Furthermore, the process of performing delayed supplementary loading on the enhancement layer data to be loaded and merging and updating it with the core layer data includes:

[0087] S401. Based on the hierarchical loading state and the scheduling priority parameter, construct a segmented loading sequence for the enhancement layer data in the waiting-to-load state, and assign weights to the segmented loading sequence according to the scheduling priority parameter to form a batch loading strategy for the enhancement layer data. The batch loading strategy is used to control the loading order and loading granularity of the enhancement layer data in the time slice dimension.

[0088] S402. During the data loading process of the enhancement layer, the enhancement layer data is fused with the core layer data of the corresponding time slice, and the data representation of the time slice is iteratively updated based on the fusion result to form a gradually evolving data representation state; the fusion process includes weighted fusion based on hierarchical weights or layered replacement processing based on data coverage relationship, and the image integrity state is synchronously adjusted during the update process;

[0089] The fusion process further includes:

[0090] The core layer data is represented as a first feature vector, and the enhancement layer data is represented as a second feature vector. A hierarchical weight function is constructed in a unified feature space.

[0091] Based on the hierarchical weight function, the corresponding weight coefficients are determined, and the first feature vector and the second feature vector are weighted and combined according to the weight coefficients to obtain the fused feature vector;

[0092] Based on the difference distribution between the fused feature vector and the first feature vector, a data overlay mask is constructed, and a hierarchical replacement operation is performed on the fused feature vector based on the data overlay mask to update the data representation of the corresponding time slice.

[0093] Furthermore, S4 takes over the resource allocation status and scheduling priority parameters generated by S3, and expands them in conjunction with the hierarchical loading status set and the prefetch order maintained in S2; specifically, the processing objects of S4 are limited to those that satisfy... The collection of time slices, in which the time slices have completed the loading of core layer data and are ready for playback. However, its enhancement layer data has not yet been fully loaded, so it enters the delayed supplementary loading stage; furthermore, S4 does not change the prefetching order already formed in S2, but under the constraint of this order, it controls the loading process of enhancement layer data in segments by scheduling priority parameters, so that the enhancement layer data forms a loading path that is progressively advanced according to priority in the time segmentation dimension, thereby completing the gradual supplementation of data details while maintaining the continuity of playback.

[0094] In this embodiment, S401 is based on the hierarchical loading state set. and scheduling priority parameters For enhancement layer data in a pending loading state, a segmented loading sequence is constructed; specifically, for any condition satisfying... Time-slicing, its enhancement layer data In S1, the data set corresponding to the detail residual coding has been defined, and in S3, the corresponding resource allocation and priority parameters have been obtained. Based on this, the enhancement layer data can be... Divide into several sub-segments along the data volume dimension. Where k represents the segment index; furthermore, the segmented loading sequence is not a simple equal division, but rather establishes a mapping relationship with the scheduling priority parameter, so that the time slice with higher priority receives a higher proportion of pre-loading share in its enhancement layer data; in some implementations, a segment weight allocation function can be constructed based on the scheduling priority parameter. Let the loading weight corresponding to the k-th segment of the enhancement layer data in the i-th time slice be represented as: ,in This represents the loading weight of the k-th enhancement layer data in the i-th time slice. This represents the segmented weight allocation function; the segmented weight allocation function is used to form a segmented loading sequence driven by priority parameters within the same time slice, and to correlate it with the prefetch order. These factors work together in the loading and scheduling process of the enhancement layer data. The determination of the number of segments, the division of segment boundaries, and the specific form of the weight function are all well-known to those skilled in the art, and will not be elaborated upon here. Through the above processing, S401 outputs a batch loading sequence of the enhancement layer constrained by the time-segmented order and segment weights, providing progressive input for subsequent fusion updates.

[0095] In this embodiment, S402 involves fusing the enhancement layer data with the core layer data of the corresponding time slice during the enhancement layer data loading process, and iteratively updating the data representation of the time slice based on the fusion result. Specifically, the inputs of S402 include core layer data, enhancement layer data segments, hierarchical loading status, and scheduling priority parameters. The processing result is not completed all at once, but rather the data representation of the time slice is updated once after each segment is loaded, so that the data representation of the time slice evolves gradually along the loading process. Furthermore, this evolution process is reflected in the gradual increase of the enhancement layer occupancy in the hierarchical loading status, and the gradual improvement of the fineness of the corresponding time slice data representation. Regarding the ratio conversion, threshold judgment, and state synchronization in the hierarchical loading status update process, the calculation or processing method is well known to those skilled in the art and will not be elaborated on here.

[0096] In some implementations, the fusion process employs a computational structure that combines vector representation in a unified feature space with hierarchical replacement; specifically, the core layer data is represented as a first feature vector. The enhancement layer data is segmented and represented as a second feature vector. Furthermore, a hierarchical weight function is constructed in a unified feature space to enable the two types of data to participate in fusion operations in the same computational space. In this embodiment, the hierarchical weight function is used to map the scheduling priority parameter and the hierarchical loading state into fusion weight coefficients, denoted as... Then the feature vectors are fused. Represented as ,in The layer weight coefficient is represented; the layer weight coefficient changes with the loading progress of the enhancement layer and the scheduling priority parameter, so that the fusion process has dynamics consistent with the loading progress; the construction method of feature vector, the selection of feature space and the specific form of weight function are well known to those skilled in the art, and will not be elaborated on here.

[0097] In this embodiment, after calculating the fused feature vector, to avoid local discontinuities caused by the overall coverage of the core layer data by the enhancement layer data, a data coverage mask construction process based on differential distribution is introduced; specifically, this is achieved by comparing the fused feature vector with the core layer feature vector. The difference distribution between them generates a data overlay mask. Represented as: ,in The difference mapping function is used to divide the fused features into regions based on the difference magnitude, local change characteristics, or distribution density, thereby determining which regions will undergo enhancement layer replacement and which regions will retain core layer data. The specific construction method and threshold setting method of the difference mapping function are well known to those skilled in the art, and will not be elaborated further here.

[0098] In some implementations, the fused feature vector is based on the data overlay mask. Perform a hierarchical replacement operation to update the data representation of the corresponding time slice; specifically, the updated time slice data representation Represented as , where ⊙ represents element-wise multiplication. This represents the updated time-slice data representation; thus, a computational chain of "weighted combination - difference detection - mask replacement" is formed during the fusion process, enabling the enhancement layer data to participate in the update in a regional manner within the time slice, rather than being completely covered, thereby maintaining the continuity and controllability of the data representation during the segmented loading process.

[0099] In this embodiment, S4, through the continuous processing of S401 and S402, completes the entire process of enhancement layer data from the pending loading state to segmented loading and then to fusion update. The segmented loading sequence ensures the orderly entry of enhancement layer data in both the time and segment dimensions; the hierarchical weight function ensures that the participation ratio and scheduling priority parameters of enhancement layer data and core layer data are consistent during the fusion process; and the data overlay mask ensures selective replacement in the spatial distribution of the fusion update process. Furthermore, as enhancement layer data in each time slice is gradually loaded, its hierarchical loading state... In The system continuously updates, and when the preset completion conditions are met, the corresponding time segment exits the enhancement layer waiting set, thereby providing updated state input for the dynamic adjustment based on playback behavior in S5. This makes the entire method form a closed loop connection in the three levels of time sequence structure, resource allocation structure and data representation structure.

[0100] Please see the appendix Figure 4 As shown, attached Figure 4 The progressive loading process of core layer data and enhancement layer data is illustrated. In this embodiment, for the same time segment, its core layer data is loaded first to form the basic playback data state. Subsequently, enhancement layer data is gradually supplemented to the corresponding time segment according to the segmented loading sequence. During the enhancement layer data loading process, the enhancement layer data is fused with the already loaded core layer data, and the data representation of the time segment is iteratively updated, so that the data state of the time segment evolves gradually along the scheduling phase. As the proportion of enhancement layer data loading increases, the data representation of the time segment transitions from the basic playback state to a high-completeness state, thereby realizing the progressive update of data expression during playback.

[0101] S5. During the replay of the event, based on the user's playback behavior information and the basic playback status, hierarchical loading status, and resource allocation status, the prefetch window range, time slice prefetch order, and scheduling priority parameters are adjusted in real time, and the hierarchical prefetch scheduling and delayed supplementary loading are continuously executed to achieve scheduling control of gradually improving the integrity of the replay image during playback.

[0102] Furthermore, the user playback behavior information includes at least one of drag operations, pause duration, and replay frequency, and the scheduling priority parameter is modified based on the user playback behavior information so that the user playback behavior information participates in the time-slicing scheduling priority calculation process.

[0103] Specifically, S2 has completed the sorting of time slices in the time distance function space and the priority loading of core layer data; S3 has completed the quantitative expression of resource supply and demand relationship and hierarchical resource allocation; and S4 has completed the segmented loading and fusion update processing of enhancement layer data. Therefore, when entering S5, the system has a stable slice sequence structure, state structure and resource allocation structure. The processing goal of S5 is no longer static scheduling, but adaptive correction of the above structure in response to dynamic behavior changes during playback, so that the scheduling process can be continuously adjusted according to changes in user behavior.

[0104] In this embodiment, the user playback behavior information includes at least one of drag operations, pause duration, and replay frequency. Drag operations characterize the user's position jump behavior on the playback timeline, pause duration characterizes the current pause state of the playback process, and replay frequency characterizes the user's repeated access characteristics within a local time interval. Furthermore, the above-mentioned behavior information can be obtained through player interaction events or playback logs and converted into behavior-related data corresponding to the time segment index. Regarding the collection method, event triggering mechanism, and data recording method of the behavior information, the calculation or processing method is well-known to those skilled in the art and will not be elaborated upon here. By introducing user playback behavior information into the scheduling process, S5 extends the original scheduling model based on resource constraints and time distance into a multi-dimensional scheduling model that includes user behavior factors.

[0105] In some implementations, the scheduling priority parameters are modified based on the user playback behavior information, allowing the user behavior information to participate in the scheduling priority calculation process of time slices. Specifically, when a change in the target playback position corresponding to a drag operation is detected, the target time slice position on which the original prefetch order depends is updated, thereby triggering the reconstruction process of the time distance function space in S2, and regenerating the prefetch candidate set and prefetch order under the new target time slice position. When a pause behavior is detected and the pause duration exceeds a preset threshold, the scheduling process no longer prioritizes pushing the time axis forward, but instead shifts the focus of resource allocation to the enhancement layer data near the current playback position, thereby supplementing the hierarchical loading state of the local time slice during the pause phase. When a continuous increase in the replay frequency is detected within a certain time interval, the scheduling priority parameter of the corresponding time slice is increased based on the original resource allocation, so that the enhancement layer data within that time interval receives a higher loading weight in subsequent scheduling cycles. It should be noted that the mapping process of behavioral information to scheduling priority parameters can be achieved through weight correction, incremental superposition, or normalization redistribution. These calculation or processing methods are well known to those skilled in the art, and will not be elaborated on here.

[0106] In this embodiment, after completing the scheduling priority parameter correction, S5 further adjusts the prefetch window range, time slice prefetch order, and resource allocation status synchronously. The adjustment of the prefetch window range is reflected in the expansion or contraction of the window boundary as the target playback position changes or user behavior changes, so that the prefetch range is always distributed around the currently interested area. The adjustment of the time slice prefetch order is reflected in the rearrangement of the time slice order in the prefetch candidate set under the influence of the new target time slice position or behavior weight. The adjustment of the resource allocation status is reflected in the recalculation of the resource allocation ratio of each time slice and its hierarchical data under the constraints of the corrected scheduling priority parameter. The above three adjustments are not executed independently, but are updated collaboratively within a unified scheduling cycle, thereby maintaining the consistency between the time sequence structure, resource allocation structure, and status structure.

[0107] In some implementations, after completing the above adjustments, S5 continues to execute the layered prefetch scheduling and enhancement layer delay supplementary loading processes from S2 to S4, so that the updated prefetch order, scheduling priority parameters, and layer loading state set continue to participate in the calculation in subsequent scheduling cycles. Specifically, S5 does not form a new processing chain on its own, but is embedded in the entire scheduling process as a closed-loop control link. By continuously introducing user behavior information, it corrects the existing scheduling structure, so that the system forms a cyclic processing mechanism of "prefetch-allocation-loading-fusion-rescheduling" during playback. In this cycle, the basic playback state set maintains the constraints on playable conditions, the layer loading state set continuously records the loading progress of layered data, the resource allocation state continuously reflects the current resource supply and demand relationship, and the scheduling priority parameters change dynamically under the combined effect of user behavior and resource status.

[0108] In this embodiment, through this step, the entire method forms a dynamic scheduling closed loop driven by user behavior in the time-slicing dimension. Within each scheduling cycle, this closed loop updates the prefetch window range, time-slicing order, and layered loading ratio based on the latest playback position, resource status, and user behavior, ensuring that data loading during playback remains consistent with playback behavior. Furthermore, during playback, as core layer data continuously ensures playback continuity, enhancement layer data is gradually added to the data representation structure of each time slice in multiple scheduling cycles. This gradually increases the enhancement layer's occupancy in the layered loading state set, completing a continuous transition from basic data representation to high-completeness data representation at the state level. Thus, this step logically forms a complete closed-loop connection with the aforementioned steps.

[0109] In some implementations, the prefetch window range is dynamically adjusted according to the current playback status, including: determining the time span parameter of the prefetch window based on the current playback rate and buffer usage status, and expanding or shrinking the prefetch window range according to the time span parameter.

[0110] Furthermore, the prefetch window range is dynamically adjusted according to the current playback state. Specifically, the current playback state includes at least the current playback rate and cache occupancy status. The playback rate characterizes the rate at which data is consumed per unit time, and the cache occupancy status characterizes the remaining capacity available on the terminal side to carry prefetched data. In this embodiment, a time span parameter is constructed based on the playback rate and cache occupancy status to characterize the time range that the prefetch window can cover along the playback time axis under the current resource conditions. The time span parameter is positively correlated with the playback rate and constrained by the cache occupancy level, thereby expanding the prefetch window coverage range when the playback rate increases and shrinking the prefetch window coverage range when the cache occupancy is close to the upper limit. Furthermore, the prefetch window range is boundary adjusted according to the time span parameter so that the prefetch window is always distributed around the current playback position and matches the resource supply and demand status. The specific calculation method and boundary constraint method of the time span parameter are well known to those skilled in the art and will not be elaborated further here.

[0111] In some implementations, the time distance of the time slice relative to the current playback position is quantified by a time distance function, which is used to establish the priority order relationship of the time slices in the scheduling process and participate in the generation process of the scheduling priority parameter.

[0112] In some implementations, the time distance between the time slice and the current playback position is quantified using a preset time distance function. Specifically, the time distance function generates a corresponding distance representation based on the relative relationship between the position of the time slice on the playback timeline and the current playback position, and maps the distance representation to an order parameter participating in the scheduling calculation. In this embodiment, the time distance function is constructed under a unified time coordinate system, enabling each time slice to have a comparable distance metric during the scheduling process, thereby forming a priority order relationship of the time slices based on the distance size. Furthermore, the output of the time distance function participates in the generation process of the scheduling priority parameter, making the time position factor participate in the calculation as a constraint condition during resource allocation and hierarchical loading. The specific form and mapping method of the time distance function are well known to those skilled in the art and will not be elaborated further here.

[0113] In some implementations, the core layer data and the enhancement layer data are stored in different cache areas, and data access scheduling processing is performed on the cache areas based on the layer loading state, including:

[0114] The core layer data is mapped to the first cache space, the enhancement layer data is mapped to the second cache space, and corresponding access weights are assigned to the first cache space and the second cache space respectively.

[0115] A cache access weight function is constructed based on the hierarchical loading state. The cache access weight function is used to quantify and allocate the access frequency of the first cache space and the second cache space according to the hierarchical loading state, and to constrain the cache read order according to the access weight, so that the core layer data and the enhancement layer data form a hierarchical access order relationship during the cache access process.

[0116] Furthermore, in some implementations, the core layer data and enhancement layer data are stored in different cache areas to form two types of data access spaces that are independent of each other in terms of storage structure. Specifically, the core layer data is mapped to a first cache space, and the enhancement layer data is mapped to a second cache space. The first cache space is used to carry the data set that constitutes the basic playable data, and the second cache space is used to carry the data set that is used for detail supplementation, thereby distinguishing the access paths of basic playback data and enhanced supplementary data in the cache structure. In this embodiment, in order to keep the cache access process consistent with the aforementioned hierarchical loading state set, access weights are assigned to the first cache space and the second cache space respectively, and a cache access weight function is constructed based on the hierarchical loading state so that the cache access process can reflect the difference in loading progress of each time slice at the core layer and enhancement layer levels. Specifically, when the core layer data of a certain time slice has not yet met the basic requirements, the cache access weight function is used to carry the data set that constitutes the basic playable data, and the cache access weight function is used to carry the data set that is used for detail supplementation, thereby distinguishing the access paths of basic playback data and enhanced supplementary data in the cache structure. When the basic playback conditions are met, the corresponding access weight is preferentially allocated to the first cache space to increase the reading frequency of core layer data. When the core layer data has met the basic playback conditions and the enhancement layer data is in a pending loading state, its corresponding access weight is transferred to the second cache space to participate in the supplementary reading of enhancement layer data. Furthermore, the cache access weight function is used to quantify and allocate the access frequency of the two types of cache spaces according to the relationship between the core layer loading occupancy and the enhancement layer loading occupancy in the hierarchical loading state, and on this basis, impose constraints on the cache reading order so that the cache reading process forms an ordered access relationship in the time slicing dimension and the hierarchical dimension, thereby enabling the core layer data to participate in the basic playback process first, and the enhancement layer data to participate in subsequent supplementary processing after meeting the conditions. It should be noted that the specific partitioning method, mapping method and access weight update method of the cache space are implementation methods well known to those skilled in the art, and will not be elaborated on here.

[0117] Based on the description of the above embodiments of the event replay data scheduling method based on time-slice prefetching, this application also discloses an event replay data scheduling system based on time-slice prefetching. This system can be a computer program (including program code) that runs the aforementioned event replay data scheduling method based on time-slice prefetching. Please see the appendix. Figure 6 As shown, the event replay data scheduling system based on time-sliced ​​prefetching can operate the following units:

[0118] The acquisition unit 110 is used to acquire the event replay source data, divide the source data along the playback time axis to obtain multiple time slice data arranged in chronological order; perform layered encoding on each time slice to generate core layer data and enhancement layer data, and establish a mapping relationship between time slices, layered data and playback time position; construct a sliced ​​layered data set based on the mapping relationship, and establish a basic playback state to represent playback continuity and a layered loading state to represent the loading progress of layered data;

[0119] The prefetch scheduling unit 120 is used to determine the time slices within the prefetch window range according to the target time slice position corresponding to the current playback request, and to determine the prefetch order according to the time distance of each time slice relative to the target time slice position; based on the slice-level data set, it performs hierarchical prefetch scheduling according to the prefetch order, prioritizes loading core layer data to establish a continuous playable data sequence, updates the basic playback state to a playable state, and marks the enhancement layer data as pending loading.

[0120] The resource allocation unit 130 is used to acquire network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Based on the information and the time distance of each time segment relative to the current playback position, it determines the resource allocation status of the time segment and its core layer data and enhancement layer data, and generates scheduling priority parameters. According to the scheduling priority parameters, it dynamically adjusts the loading ratio of core layer data and enhancement layer data of the time segment within the prefetch window, and updates the layer loading status.

[0121] The enhancement loading unit 140 is used to perform delayed supplementary loading of the enhancement layer data to be loaded based on the layer loading status and the scheduling priority parameter; during the loading process, the enhancement layer data is loaded step by step according to the time slice order and the scheduling priority parameter, and is merged and updated with the already loaded core layer data to improve the picture integrity status of the corresponding time slice.

[0122] The dynamic scheduling unit 150 is used to adjust the prefetch window range, time slice prefetch order and scheduling priority parameters in real time according to user playback behavior information and the basic playback state, hierarchical loading state and resource allocation state during the playback of the event replay, and continuously execute the hierarchical prefetch scheduling and delayed supplementary loading to realize the scheduling control of gradually improving the integrity of the replay screen during the playback process.

[0123] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for scheduling event replay data based on time-sliced ​​prefetching, characterized in that, Includes the following steps: S1. Obtain the event replay source data, and divide the source data along the playback time axis to obtain multiple time segment data arranged in time order; Each time segment is encoded in layers to generate core layer data and enhancement layer data, and a mapping relationship is established between time segments, layer data and playback time position. Based on the mapping relationship, a segmented hierarchical data set is constructed, and a basic playback state for characterizing playback continuity and a hierarchical loading state for characterizing the loading progress of hierarchical data are established. S2. Based on the target time segment position corresponding to the current playback request, determine the time segments within the prefetch window range, and determine the prefetch order based on the time distance of each time segment relative to the target time segment position; Based on the segmented hierarchical data set, hierarchical prefetching scheduling is performed according to the prefetching order. Core layer data is loaded first to establish a continuous playable data sequence, and the basic playback state is updated to playable state. At the same time, the enhancement layer data is marked as pending loading. S3. Obtain network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Based on the information and the time distance of each time segment relative to the current playback position, determine the resource allocation status corresponding to the time segment and its core layer data and enhancement layer data, and generate scheduling priority parameters. According to the scheduling priority parameters, dynamically adjust the loading ratio of core layer data and enhancement layer data of the time segment within the prefetch window, and update the layer loading status. S4. Based on the layer loading status and the scheduling priority parameter, perform delayed supplementary loading of the enhancement layer data to be loaded; during the loading process, gradually load the enhancement layer data according to the time slice order and the scheduling priority parameter, and merge and update it with the already loaded core layer data to improve the picture integrity status of the corresponding time slice. S5. During the replay of the event, based on the user's playback behavior information and the basic playback status, hierarchical loading status, and resource allocation status, the prefetch window range, time slice prefetch order, and scheduling priority parameters are adjusted in real time, and the hierarchical prefetch scheduling and delayed supplementary loading are continuously executed to achieve scheduling control of gradually improving the integrity of the replay image during playback.

2. The event replay data scheduling method based on time-slicing prefetching according to claim 1, characterized in that, The step of determining multiple time segments within a prefetch window based on the target time segment position corresponding to the current playback request, and determining the prefetch order based on the time distance of each time segment relative to the target time segment position, includes: S201. Based on the target time segment position, construct a time distance function space along the playback time axis. Using the target time segment as a reference point, express the time segments on the time axis by distance quantization. In the time distance function space, determine the prefetch window range according to a preset window function, and select multiple time segments within the prefetch window range to form a prefetch candidate set. S202. Based on the time distance function space, assign a corresponding time distance weight to each time slice in the prefetch candidate set, and construct a priority sequence of time slices according to the time distance weight; perform sorting processing on the prefetch candidate set according to the priority sequence to generate a prefetch order that satisfies the time continuity constraint, wherein the time distance weight has a monotonic mapping relationship with the time distance of the time slice relative to the target time slice position.

3. The event replay data scheduling method based on time-slicing prefetching according to claim 1, characterized in that, The process of acquiring network bandwidth status information, terminal cache occupancy status information, and current playback rate information, and determining resource allocation status and scheduling priority parameters, includes: S301. Perform bandwidth time-series modeling processing on the network bandwidth status information to obtain a bandwidth change sequence; perform cache capacity status modeling processing on the terminal cache occupancy status information to obtain a cache occupancy sequence; and construct a data consumption rate model by combining the current playback rate information; based on the bandwidth change sequence, the cache occupancy sequence, and the data consumption rate model, generate a resource status vector to characterize the supply and demand relationship of system resources. S302. Based on the resource state vector and the position distribution of each time slice in the time distance function space, construct a multi-dimensional weight allocation model for the core layer data and enhancement layer data of each time slice; according to the multi-dimensional weight allocation model, map system resources to each time slice and its hierarchical data, generate corresponding resource allocation states, and determine the scheduling priority parameters of the hierarchical data based on the resource allocation states, wherein the multi-dimensional weight allocation model includes at least the combination relationship of time distance weight, cache occupancy weight and bandwidth allocation weight.

4. The event replay data scheduling method based on time-slicing prefetching according to claim 1, characterized in that, The process of performing delayed supplementary loading on the enhancement layer data to be loaded and merging and updating it with the core layer data includes: S401. Based on the hierarchical loading state and the scheduling priority parameter, construct a segmented loading sequence for the enhancement layer data in the waiting-to-load state, and assign weights to the segmented loading sequence according to the scheduling priority parameter to form a batch loading strategy for the enhancement layer data. The batch loading strategy is used to control the loading order and loading granularity of the enhancement layer data in the time slice dimension. S402. During the data loading process of the enhancement layer, the enhancement layer data is fused with the core layer data of the corresponding time slice, and the data representation of the time slice is iteratively updated based on the fusion result to form a gradually evolving data representation state; the fusion process includes weighted fusion based on hierarchical weights or layered replacement processing based on data coverage relationship, and the image integrity state is synchronously adjusted during the update process.

5. The method for scheduling event replay data based on time-slicing prefetching according to claim 1, characterized in that, The prefetch window range is dynamically adjusted according to the current playback status, including: determining the time span parameter of the prefetch window based on the current playback rate and cache usage status, and expanding or shrinking the prefetch window range according to the time span parameter.

6. The method for scheduling event replay data based on time-slicing prefetching according to claim 1, characterized in that, The time distance between the time slice and the current playback position is quantified and calculated using a time distance function. This time distance function is used to establish the priority order relationship of the time slices during the scheduling process and participates in the generation process of the scheduling priority parameters.

7. The event replay data scheduling method based on time-slicing prefetching according to claim 1, characterized in that, The user playback behavior information includes at least one of drag operations, pause duration, and replay frequency. The scheduling priority parameter is modified based on the user playback behavior information so that the user playback behavior information participates in the time-slicing scheduling priority calculation process.

8. The method for scheduling event replay data based on time-slicing prefetching according to claim 1, characterized in that, The core layer data and enhancement layer data are stored in different cache areas, and data access scheduling processing is performed on the cache areas based on the layer loading status, including: The core layer data is mapped to the first cache space, the enhancement layer data is mapped to the second cache space, and corresponding access weights are assigned to the first cache space and the second cache space respectively. A cache access weight function is constructed based on the hierarchical loading state. The cache access weight function is used to quantify and allocate the access frequency of the first cache space and the second cache space according to the hierarchical loading state, and to constrain the cache read order according to the access weight, so that the core layer data and the enhancement layer data form a hierarchical access order relationship during the cache access process.

9. A method for scheduling event replay data based on time-slicing prefetching according to claim 4, characterized in that, The fusion process includes: The core layer data is represented as a first feature vector, and the enhancement layer data is represented as a second feature vector. A hierarchical weight function is constructed in a unified feature space. Based on the hierarchical weight function, the corresponding weight coefficients are determined, and the first feature vector and the second feature vector are weighted and combined according to the weight coefficients to obtain the fused feature vector; Based on the difference distribution between the fused feature vector and the first feature vector, a data overlay mask is constructed, and a hierarchical replacement operation is performed on the fused feature vector based on the data overlay mask to update the data representation of the corresponding time slice.

10. A sports replay data scheduling system based on time-sliced ​​prefetching, characterized in that, The system includes: The acquisition unit is used to acquire the source data of the event replay, divide the source data along the playback time axis to obtain multiple time slice data arranged in chronological order; perform hierarchical encoding on each time slice to generate core layer data and enhancement layer data, and establish a mapping relationship between time slices, hierarchical data and playback time position; construct a slice hierarchical data set based on the mapping relationship, and establish a basic playback state to represent playback continuity and a hierarchical loading state to represent the loading progress of hierarchical data. The prefetch scheduling unit is used to determine the time slices within the prefetch window range according to the target time slice position corresponding to the current playback request, and to determine the prefetch order according to the time distance of each time slice relative to the target time slice position; based on the slice-level data set, it performs hierarchical prefetch scheduling according to the prefetch order, prioritizes loading core layer data to establish a continuous playable data sequence, updates the basic playback state to a playable state, and marks the enhancement layer data as pending loading. The resource allocation unit is used to acquire network bandwidth status information, terminal cache occupancy status information, and current playback rate information. Based on the information and the time distance of each time segment relative to the current playback position, it determines the resource allocation status of the time segment and its core layer data and enhancement layer data, and generates scheduling priority parameters. According to the scheduling priority parameters, it dynamically adjusts the loading ratio of core layer data and enhancement layer data of the time segment within the prefetch window, and updates the layer loading status. The enhancement loading unit is used to perform delayed supplementary loading of the enhancement layer data to be loaded based on the layer loading status and the scheduling priority parameter; during the loading process, the enhancement layer data is loaded step by step according to the time slice order and the scheduling priority parameter, and merged and updated with the already loaded core layer data to improve the picture integrity status of the corresponding time slice. The dynamic scheduling unit is used to adjust the prefetch window range, time slice prefetch order and scheduling priority parameters in real time according to user playback behavior information, playback base status, hierarchical loading status and resource allocation status during the replay of the event. It also continuously executes the hierarchical prefetch scheduling and delayed supplementary loading to achieve scheduling control that gradually improves the integrity of the replay image during the playback process.

Images