A film review system that supports multi-track customized playback and version switching

Abstract

This invention relates to the field of film and television review systems, and discloses a film and television review system that supports multi-track customized playback and version switching. The system includes a multi-track management module, a playback customization module, a version switching module, a user interaction module, and a performance optimization module. By setting a playback customization engine, when performing multi-track combined playback, dynamic parameter adjustment rules are generated based on user preference data, and independent adjustment thresholds are set for different track types to improve the customization compatibility of video, audio, and subtitle tracks. At the same time, artificial intelligence algorithms are deeply integrated into the playback parameter calculation process to analyze the correlation between user preference feature vectors and track collaborative feature data in real time, and dynamically adjust playback speed, volume balance, and subtitle synchronization parameters. This solves the problem of playback sequence interruption caused by the inability to dynamically customize playback parameters in traditional systems, and improves the efficiency of the review process.

Description

Technical Field

[0001] This invention relates to the field of film and television review systems, specifically a film and television review system that supports multi-track customized playback and version switching. Background Technology

[0002] Film review is a core step in the process of newsreel production, from reporter editing to final broadcast. This process includes material editing, dubbing, content review, and broadcast scheduling. With industry development, this process has expanded to film, animation, and advertising media, forming a professional collaborative model. In the film and television production review stage, existing technologies primarily rely on traditional review systems and tools. These systems typically support basic video playback and pause functions, allowing reviewers to examine the film frame by frame. Some systems also attempt to provide version management functions to record and manage different versions during film and television production. Furthermore, some review tools can display individual effects from compositing, lighting, special effects, animation, and editing stages of a film or television work, but comprehensive integration and intuitive comparison are not yet achieved.

[0003] Currently, due to the massive heterogeneous data sources involved in film and television review processes, traditional linear playback systems can only load preset track combinations when playing multi-track content. They cannot dynamically customize playback parameters based on user preferences. When a user needs to temporarily adjust the parameters of a track, the entire playback sequence will be interrupted and reassembled, resulting in a decrease in the efficiency of the review process. At the same time, when switching between multiple versions of content, the review system cannot intelligently identify the spatiotemporal correlation between the current playback content and the target version. Scene frame offsets cause audio-visual asynchrony and special effects misalignment after switching, and the switching process lacks an automated correction mechanism. In cross-track collaborative review, due to the huge differences in the real-time processing requirements of different track data, it is impossible to achieve hierarchical scheduling of multi-track resources, causing high-priority tracks to be blocked by low-priority tasks, further affecting the smoothness of multi-track customized playback and the accuracy of version switching.

[0004] Therefore, a film review system that supports multi-track customized playback and version switching is proposed to solve the above problems. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a film review system that supports multi-track customized playback and version switching, thus solving the problems mentioned in the background section.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a film review system that supports multi-track customized playback and version switching, the system including a multi-track management module, a playback customization module, a version switching module and a user interaction module;

[0009] The multi-track management module collects and stores multi-track data of film and television content, including video tracks, audio tracks, subtitle tracks, and special effects tracks;

[0010] The playback customization module dynamically adjusts the playback parameters of multi-track data based on user input preference data and artificial intelligence algorithms to generate a customized playback sequence.

[0011] The version switching module calls the version decision engine to perform multi-version matching processing based on the version switching request data and generates a target version switching instruction.

[0012] The user interaction module receives user input data and outputs real-time feedback of customized playback sequences and target version switching instructions. The multi-track data includes track identifiers, time axis parameters, and content feature parameters. The artificial intelligence algorithm includes a deep learning model and a collaborative filtering algorithm. The version decision engine generates switching instructions based on similarity calculations using preset version library data.

[0013] Preferably, the system also includes a performance optimization module;

[0014] The performance optimization module performs real-time resource allocation processing based on system load data and resource scheduling algorithms, and generates optimized resource allocation instructions. The resource scheduling algorithms are the minimum number of connections algorithm and the priority queue algorithm. The system load data includes CPU utilization, memory usage and network bandwidth parameters.

[0015] Preferably, the resource allocation processing steps performed by the performance optimization module are as follows:

[0016] S31, Monitor system load data;

[0017] S32. When the load exceeds the threshold, a priority queue algorithm is used to schedule the tasks of the multi-track management module, playback customization module, and version switching module.

[0018] S33. Dynamically allocate computing resources based on the scheduling results and output optimized resource allocation instructions. The preset thresholds are CPU utilization greater than 80% and memory occupancy greater than 70%. The priority is calculated based on the track data type and user priority weight.

[0019] Preferably, the multi-track management module includes a multi-track data acquisition unit and a multi-track data storage unit;

[0020] The multi-track data acquisition unit performs multi-track data acquisition and processing of film and television content through film and television editing software interfaces and API integration tools, and generates original multi-track datasets;

[0021] The multi-track data storage unit partitions and stores the original multi-track dataset based on a distributed database system, and establishes a track index table to generate and optimize multi-track data. The original multi-track dataset includes video frame rate parameters, audio sampling rate parameters, subtitle timestamp parameters, and special effects layer parameters. The distributed database system adopts a Hadoop and Spark architecture. The track index table includes track type identifiers, timeline start positions, and content feature hash values.

[0022] Preferably, the playback customization module includes a user preference analysis unit and a playback parameter adjustment unit;

[0023] The user preference analysis unit performs preference feature extraction processing based on user input preference data and a collaborative filtering algorithm to generate a user preference feature vector.

[0024] The playback parameter adjustment unit dynamically adjusts the stored optimized multi-track data based on the user preference feature vector, including playback speed parameters, volume balance parameters, and subtitle synchronization parameters, and generates a customized playback sequence. The user input preference data includes playback speed preference value, volume preference level, and subtitle language preference. The collaborative filtering algorithm constructs a preference matrix based on the user's historical behavior data, and the customized playback sequence is represented as a combination of multi-track parameters aligned with the time axis.

[0025] Preferably, the playback parameter adjustment unit performs the following steps for dynamic parameter adjustment:

[0026] S61. Constructing a Preference Matrix: Based on historical user behavior data, including playback speed preference values, volume preference levels, and subtitle language preferences, construct a user-item preference matrix. The rows represent user identifiers, and the columns represent playback parameter types, including playback speed, volume balance, and subtitle synchronization. (Matrix elements) Indicates user For parameters Preference rating;

[0027] For missing values, the average user preference is used to fill them in, and the singular value decomposition (SVD) is used to reduce the dimensionality of the preference matrix.

[0028] in For the user-item preference matrix, For the user feature matrix, It is a singular value matrix. For the project feature matrix, This is the transpose of the project feature matrix;

[0029] Extracting user feature vectors As input to the user preference feature vector;

[0030] S62. Obtain user preference feature vectors and store optimized multi-track data;

[0031] S63. Based on a deep learning model, perform track coordination analysis on storage-optimized multi-track data to generate track coordination feature data;

[0032] S64. Calculate the playback parameter adjustment weights based on the user preference feature vector and track coordination feature data, and generate the parameter adjustment weight matrix;

[0033] S65. An interpolation algorithm is used to adjust the playback speed parameters, volume balance parameters, and subtitle synchronization parameters of multi-track data in real time according to the parameter adjustment weight matrix, and output a customized playback sequence. The deep learning model is a convolutional neural network (CNN) and a long short-term memory network (LSTM). The track collaborative feature data includes video-audio synchronization coefficients and subtitle-effect alignment parameters. The interpolation algorithm is linear interpolation and spline interpolation.

[0034] Preferably, the version switching module includes a version control unit and a version decision engine unit;

[0035] The version control unit is used to store a preset multi-version dataset, establish a version feature index, and generate version control data.

[0036] The version decision engine unit performs similarity matching processing based on version switching request data and version library data to generate target version switching instructions. The multi-version dataset includes film and television content parameters of different editing versions, and the version feature index includes version identifier, content difference features and time axis offset parameters. The similarity matching processing adopts the cosine similarity algorithm and the dynamic time warping (DTW) algorithm.

[0037] Preferably, the similarity matching processing steps performed by the version decision engine unit are as follows:

[0038] S81. Obtain version switch request data and repository data;

[0039] S82. Based on the artificial intelligence bionic algorithm, search for target version features that match the version switching request data in the version database data, and generate a candidate version set;

[0040] S83. Calculate the similarity score of each version in the candidate version set: Euclidean distance of the content features of each candidate version. Time axis alignment error Perform Min-Max standardization:

[0041]

[0042]

[0043] in Euclidean distance for content features To determine the degree of difference in standardized content features, This is due to time axis alignment error. This is the standardized time axis error. and These are the minimum and maximum values ​​of the corresponding indicators in the candidate version set;

[0044] Calculate similarity score :

[0045]

[0046] in For content feature weights, , This is the time axis error weight, with a value of 0.4, and , The overall similarity score is calculated.

[0047] Select the highest-scoring version to generate the target version switching instruction.

[0048] Preferably, when the artificial intelligence biomimetic algorithm is a particle swarm optimization algorithm, the specific steps for performing the search operation are as follows:

[0049] S821. Initialize the particle swarm position and the maximum number of iterations T, with values ​​ranging from 20 to 50. Normalize the repository search space boundary to the interval [0,1]. Set the lower bound of the original search space. and the Upper Realm After normalization: , ;

[0050] The position initialization formula is:

[0051]

[0052] in Let i be the position of particle i in the d-dimensional version library search space. The values ​​are random numbers in the interval [0,1], and the particle positions are initialized within [0,1] to ensure consistent update scales in subsequent updates.

[0053] S822, Exploration Phase: The particle moves towards the global optimal position, and the position update formula is:

[0054]

[0055] in Let i be the updated position of particle i in d-dimensional space. Let i be the global optimal position, i be the particle index, and d be the dimension of the search space;

[0056] If the new position has a higher similarity score, then the original position is replaced.

[0057] S823, Development Phase: Calculating the New Location:

[0058]

[0059] in Let t be the new position of particle i in dimension d, and t be the current iteration number. , For the upper and lower bounds of the search space,

[0060] When the similarity score improves, the position is updated;

[0061] S824. Check the termination condition: when the number of iterations reaches... And the global optimal solution satisfies the following for 3 consecutive times:

[0062]

[0063] Output the optimal solution; otherwise, return S822.

[0064] in The sequence value is the similarity score. Index for iteration count, The similarity score for the k-th iteration is the current optimal solution. For the first The similarity score of the next iteration is the historical best solution. The convergence threshold is a constant of 0.00001.

[0065] S825, Generate target version switching instructions based on the characteristics of the optimal target version.

[0066] Preferably, the user interaction module includes an input receiving unit and a feedback output unit;

[0067] The input receiving unit receives user input data through a graphical user interface (GUI) and a voice recognition interface, including playback customization preference data and version switching request data.

[0068] The feedback output unit renders the customized playback sequence and target version switching instructions into a real-time visualization interface and outputs it to the display device. The graphical user interface supports drag-and-drop track editing, and the real-time visualization interface includes multi-track waveforms and a version switching progress bar.

[0069] (III) Beneficial Effects

[0070] Compared with existing technologies, this invention provides a film review system that supports multi-track customized playback and version switching, and has the following beneficial effects:

[0071] 1. In this invention, by setting up a playback customization engine, dynamic parameter adjustment rules are generated based on user preference data when performing multi-track combined playback, and independent adjustment thresholds are set for different track types to improve the customization compatibility of video, audio, and subtitle tracks. At the same time, artificial intelligence algorithms are deeply integrated into the playback parameter calculation process to analyze the correlation between user preference feature vectors and track collaborative feature data in real time, and dynamically adjust playback speed, volume balance, and subtitle synchronization parameters. This solves the problem of playback sequence interruption caused by the inability to dynamically customize playback parameters in traditional systems, and improves the efficiency of the film review process.

[0072] 2. In this invention, by setting up a version decision engine, when switching between multiple versions of content, the system calculates the time axis offset compensation value and content feature difference between the current playback content and the target version, and identifies the risk of spatiotemporal misalignment in the version switching in real time; the system automatically triggers the special effects layer alignment correction mechanism and audio-visual synchronization compensation protocol to eliminate the problem of special effects misalignment and audio-visual asynchrony after switching caused by the lack of intelligent recognition, and ensures a high degree of consistency in the spatiotemporal correlation of each track element after switching.

[0073] 3. In this invention, by setting up a resource hierarchical scheduler, the system load status is dynamically monitored according to the preset track data type priority weight system during cross-track collaborative review; the minimum connection number algorithm is used to dynamically allocate computing resources in real time, eliminating the playback stuttering phenomenon caused by high-priority track tasks being blocked by low-priority tasks, and achieving collaborative optimization of multi-track playback smoothness and resource coordination. Attached Figure Description

[0074] Figure 1 This is a diagram illustrating the architecture of a film review system that supports multi-track customized playback and version switching, as described in this invention. Detailed Implementation

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] Please see Figure 1 This is a film review system that supports multi-track customized playback and version switching. The system includes a multi-track management module, a playback customization module, a version switching module, and a user interaction module.

[0077] The multi-track management module collects and stores multi-track data of film and television content, including video tracks, audio tracks, subtitle tracks, and special effects tracks;

[0078] The playback customization module dynamically adjusts playback parameters of multi-track data based on user input preference data and artificial intelligence algorithms to generate customized playback sequences.

[0079] The version switching module calls the version decision engine to perform multi-version matching processing based on the version switching request data and generates the target version switching instruction.

[0080] The user interaction module receives user input data and outputs real-time feedback on customized playback sequences and target version switching instructions. The multi-track data includes track identifiers, timeline parameters, and content feature parameters. The artificial intelligence algorithms include deep learning models and collaborative filtering algorithms. The version decision engine generates switching instructions based on similarity calculations using preset version library data.

[0081] The system also includes a performance optimization module;

[0082] The performance optimization module performs real-time resource allocation processing based on system load data and resource scheduling algorithms, generating optimized resource allocation instructions. The resource scheduling algorithms are the minimum number of connections algorithm and the priority queue algorithm. The system load data includes CPU utilization, memory usage, and network bandwidth parameters.

[0083] The performance optimization module performs resource allocation processing steps as follows:

[0084] S31, Monitor system load data;

[0085] S32. When the load exceeds the threshold, a priority queue algorithm is used to schedule the tasks of the multi-track management module, playback customization module, and version switching module.

[0086] S33. Dynamically allocate computing resources based on the scheduling results and output optimized resource allocation instructions. The preset thresholds are CPU utilization greater than 80% and memory occupancy greater than 70%. The priority is calculated based on the track data type and user priority weight.

[0087] The multi-track management module includes a multi-track data acquisition unit and a multi-track data storage unit;

[0088] The multitrack data acquisition unit performs multitrack data acquisition and processing of film and television content through film and television editing software interfaces and API integration tools, and generates raw multitrack datasets;

[0089] The multitrack data storage unit partitions and stores the original multitrack dataset based on a distributed database system, and establishes a track index table to generate and optimize multitrack data. The original multitrack dataset includes video frame rate parameters, audio sampling rate parameters, subtitle timestamp parameters, and special effects layer parameters. The distributed database system adopts a Hadoop and Spark architecture, and the track index table includes track type identifiers, timeline start positions, and content feature hash values.

[0090] The playback customization module includes a user preference analysis unit and a playback parameter adjustment unit;

[0091] The user preference analysis unit extracts preference features based on user input preference data and a collaborative filtering algorithm to generate user preference feature vectors.

[0092] The playback parameter adjustment unit dynamically adjusts the parameters of the stored optimized multi-track data based on the user preference feature vector, including playback speed parameters, volume balance parameters, and subtitle synchronization parameters, and generates a customized playback sequence. The user input preference data includes playback speed preference value, volume preference level, and subtitle language preference. The collaborative filtering algorithm constructs a preference matrix based on the user's historical behavior data, and the customized playback sequence is represented as a combination of multi-track parameters aligned with the time axis.

[0093] The steps for the playback parameter adjustment unit to perform dynamic parameter adjustment are as follows:

[0094] S61. Constructing a Preference Matrix: Based on historical user behavior data, including playback speed preference values, volume preference levels, and subtitle language preferences, construct a user-item preference matrix. The rows represent user identifiers, and the columns represent playback parameter types, including playback speed, volume balance, and subtitle synchronization. (Matrix elements) Indicates user For parameters Preference rating;

[0095] For missing values, the average user preference is used to fill them in, and the singular value decomposition (SVD) is used to reduce the dimensionality of the preference matrix.

[0096] in For the user-item preference matrix, For the user feature matrix, It is a singular value matrix. For the project feature matrix, This is the transpose of the project feature matrix;

[0097] Extracting user feature vectors As input to the user preference feature vector;

[0098] S62. Obtain user preference feature vectors and store optimized multi-track data;

[0099] S63. Based on a deep learning model, perform track coordination analysis on storage-optimized multi-track data to generate track coordination feature data;

[0100] S64. Calculate the playback parameter adjustment weights based on the user preference feature vector and track coordination feature data, and generate the parameter adjustment weight matrix;

[0101] S65. An interpolation algorithm is used to adjust the playback speed parameters, volume balance parameters, and subtitle synchronization parameters of multi-track data in real time according to the parameter adjustment weight matrix, and output a customized playback sequence. The deep learning model is a convolutional neural network (CNN) and a long short-term memory network (LSTM). The track collaborative feature data includes video-audio synchronization coefficients and subtitle-effect alignment parameters. The interpolation algorithm is linear interpolation and spline interpolation.

[0102] The version switching module includes a repository management unit and a version decision engine unit;

[0103] The version control unit is used to store a pre-defined multi-version dataset, build a version feature index, and generate version control data;

[0104] The version decision engine unit performs similarity matching processing based on version switching request data and version library data to generate target version switching instructions. The multi-version dataset includes film and television content parameters from different editing versions, and the version feature index includes version identifiers, content difference features, and time axis offset parameters. The similarity matching processing adopts the cosine similarity algorithm and the dynamic time warping (DTW) algorithm.

[0105] The steps for the version decision engine unit to perform similarity matching are as follows:

[0106] S81. Obtain version switch request data and repository data;

[0107] S82. Based on the artificial intelligence bionic algorithm, search for target version features that match the version switching request data in the version database data, and generate a candidate version set;

[0108] S83. Calculate the similarity score of each version in the candidate version set: Euclidean distance of the content features of each candidate version. Time axis alignment error Perform Min-Max standardization:

[0109]

[0110]

[0111] in Euclidean distance for content features To determine the degree of difference in standardized content features, This is due to time axis alignment error. This is the standardized time axis error. and These are the minimum and maximum values ​​of the corresponding indicators in the candidate version set;

[0112] Calculate similarity score :

[0113]

[0114] in For content feature weights, , This is the time axis error weight, with a value of 0.4, and , The overall similarity score is calculated.

[0115] Select the highest-scoring version to generate the target version switching instruction.

[0116] When the artificial intelligence biomimetic algorithm is the particle swarm optimization algorithm, the specific steps for performing the search operation are as follows:

[0117] S821. Initialize the particle swarm position and the maximum number of iterations T=20, normalize the repository search space boundary to the [0,1] interval, and set the lower bound of the original search space. and the Upper Realm After normalization: , ;

[0118] The position initialization formula is:

[0119]

[0120] in Let i be the position of particle i in the d-dimensional version library search space. The values ​​are random numbers in the interval [0,1], and the particle positions are initialized within [0,1] to ensure consistent update scales in subsequent updates.

[0121] S822, Exploration Phase: The particle moves towards the global optimal position, and the position update formula is:

[0122]

[0123] in Let i be the updated position of particle i in d-dimensional space. Let i be the global optimal position, i be the particle index, and d be the dimension of the search space;

[0124] If the new position has a higher similarity score, then the original position is replaced.

[0125] S823, Development Phase: Calculating the New Location:

[0126]

[0127] in Let t be the new position of particle i in dimension d, and t be the current iteration number. , For the upper and lower bounds of the search space,

[0128] When the similarity score improves, the position is updated;

[0129] S824. Check the termination condition: when the number of iterations reaches... And the global optimal solution satisfies the following for 3 consecutive times:

[0130]

[0131] Output the optimal solution; otherwise, return S822.

[0132] in The sequence value is the similarity score. Index for iteration count, The similarity score for the k-th iteration is the current optimal solution. For the first The similarity score of the next iteration is the historical best solution. The convergence threshold is a constant of 0.00001.

[0133] S825, Generate target version switching instructions based on the characteristics of the optimal target version.

[0134] The user interaction module includes an input receiving unit and a feedback output unit;

[0135] The input receiving unit receives user input data, including playback customization preference data and version switching request data, through a graphical user interface (GUI) and a voice recognition interface.

[0136] The feedback output unit renders the customized playback sequence and target version switching instructions into a real-time visualization interface and outputs it to the display device. The graphical user interface supports drag-and-drop track editing, and the real-time visualization interface includes multi-track waveforms and version switching progress bars.

[0137] Example 1: Multi-track collaborative screening in film post-production

[0138] During post-production, directors need to efficiently review high-precision multi-track combinations to optimize the playback smoothness and consistency of key segments. The system starts through a multi-track management module, utilizing an API interface to integrate professional editing software to collect raw multi-track datasets, including video frame rate parameters, audio sampling rate parameters, subtitle timestamp parameters, and special effects layer parameters. The data storage unit uses a Hadoop distributed system for partitioned storage and establishes a track index table to ensure fast data retrieval.

[0139] The playback customization module then intervenes: the user preference analysis unit parses the preference data input by the director and generates a user preference feature vector using a collaborative filtering algorithm; the playback parameter adjustment unit processes the track collaborative feature data based on the deep learning model LSTM, calculates the video-audio synchronization coefficient and subtitle-effect alignment parameters, and adjusts the playback speed, volume balance, and subtitle synchronization parameters in real time using a spline interpolation algorithm, outputting a seamlessly connected customized playback sequence. The user interaction module provides real-time visual feedback through a graphical user interface (GUI), supports drag-and-drop track editing, and allows the director to intuitively adjust the position of the effects track, with the feedback output unit immediately rendering and updating.

[0140] Significant technical benefits: The system achieves dynamic linkage of multi-track parameters, improving review efficiency and enhancing the accuracy of track collaborative analysis, thus solving the problem of playback sequence interruption caused by parameter adjustments in traditional systems.

[0141] Example 2: Review and approval of switching between international versions of multiple film and television dramas

[0142] In the international distribution review of film and television dramas, producers need to seamlessly switch between the director's cut and the international version to ensure the spatiotemporal consistency of all track elements and avoid audio-visual desynchronization and special effects misalignment. The version switching module's main process is as follows: the version repository management unit pre-stores multiple version datasets and generates version repository data; after receiving the switching request data, the version decision engine unit uses the Particle Swarm Optimization (PSO) algorithm for matching processing and initializes the particle swarm positions. Where l=0, u=1, r=0.83; Update position during exploration phase: Calculate the new location during the development phase: t represents the current iteration number, and the final output is the target version switching instruction with the highest similarity score.

[0143] The playback customization module performs dynamic adjustments synchronously, aligning the subtitle track timeline offset based on the DTW algorithm; the performance optimization module monitors system load data, and when it exceeds a threshold, the priority queue algorithm schedules tasks to independent nodes to optimize resource allocation; the user interaction module displays a version switching progress bar through a real-time visual interface to ensure transparent operation.

[0144] Significant technical benefits: Improved accuracy of version switching, reduced resource conflict rate, elimination of special effects deviations caused by time and space misalignment, and shortened switching response time.

[0145] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A film review system supporting multi-track customized playback and version switching, characterized in that: The system includes a multi-track management module, a playback customization module, a version switching module, a user interaction module, and a performance optimization module; The multi-track management module collects and stores multi-track data of film and television content, including video tracks, audio tracks, subtitle tracks, and special effects tracks; The playback customization module dynamically adjusts the playback parameters of multi-track data based on user input preference data and artificial intelligence algorithms to generate a customized playback sequence. The version switching module calls the version decision engine to perform multi-version matching processing based on the version switching request data and generates a target version switching instruction. The user interaction module receives user input data and outputs real-time feedback of customized playback sequence and target version switching instructions. The multi-track data includes track identifiers, time axis parameters and content feature parameters. The artificial intelligence algorithm includes deep learning model and collaborative filtering algorithm. The version decision engine generates switching instructions based on similarity calculation of preset version library data. The performance optimization module performs real-time resource allocation processing based on system load data and resource scheduling algorithms, and generates optimized resource allocation instructions. The resource scheduling algorithms are the minimum number of connections algorithm and the priority queue algorithm. The system load data includes CPU utilization, memory usage and network bandwidth parameters. The version switching module includes a version repository management unit and a version decision engine unit; The version control unit is used to store a preset multi-version dataset, establish a version feature index, and generate version control data. The version decision engine unit performs similarity matching processing based on version switching request data and version library data to generate target version switching instructions. The multi-version dataset includes film and television content parameters of different editing versions, and the version feature index includes version identifier, content difference features and time axis offset parameters. The similarity matching processing adopts cosine similarity algorithm and dynamic time warping (DTW) algorithm. The similarity matching process performed by the version decision engine unit is as follows: S81. Obtain version switch request data and repository data; S82. Based on the artificial intelligence bionic algorithm, search for target version features that match the version switching request data in the version database data, and generate a candidate version set; S83. Calculate the similarity score of each version in the candidate version set: Euclidean distance of the content features of each candidate version. Time axis alignment error Perform Min-Max standardization: in Euclidean distance for content features To determine the degree of difference in standardized content features, This is due to time axis alignment error. This is the standardized time axis error. and These are the minimum and maximum values ​​of the corresponding indicators in the candidate version set; Calculate similarity score : in For content feature weights, , This is the time axis error weight, with a value of 0.4, and , The overall similarity score is calculated. Select the highest-scoring version to generate the target version switching instruction; When the artificial intelligence biomimetic algorithm is a particle swarm optimization algorithm, the specific steps for performing the search operation are as follows: S821. Initialize the particle swarm position and the maximum number of iterations T=20, normalize the repository search space boundary to the [0,1] interval, and set the lower bound of the original search space. and the Upper Realm After normalization: , ; The position initialization formula is: Among them Let i be the position of particle i in the d-dimensional version library search space. The values ​​are random numbers in the interval [0,1], and the particle positions are initialized within [0,1] to ensure consistent update scales in subsequent updates. S822, Exploration Phase: The particle moves towards the global optimal position, and the position update formula is: in Let be the updated position of particle i in d-dimensional space. The global optimal position is represented by i, where i is the particle index and d is the dimension of the search space. If the new position has a higher similarity score, then the original position is replaced. S823, Development Phase: Development Phase: Calculating New Location: in Let t be the new position of particle i in dimension d, and t be the current iteration number. , For the upper and lower bounds of the search space, When the similarity score improves, the position is updated; S824. Check the termination condition: When the number of iterations reaches... And the global optimal solution satisfies the following for 3 consecutive times: Output the optimal solution; otherwise, return S822. in The sequence value is the similarity score. Index for iteration count, The similarity score for the k-th iteration is the current optimal solution. For the first The similarity score of the next iteration is the historical best solution. The convergence threshold is a constant of 0.00001. S825, Generate target version switching instructions based on the characteristics of the optimal target version.

2. The film review system supporting multi-track customized playback and version switching according to claim 1, characterized in that: The performance optimization module performs resource allocation processing steps as follows: S31, Monitor system load data; S32. When the load exceeds the threshold, a priority queue algorithm is used to schedule the tasks of the multi-track management module, playback customization module, and version switching module. S33. Dynamically allocate computing resources based on the scheduling results and output optimized resource allocation instructions. The preset thresholds are CPU utilization greater than 80% and memory occupancy greater than 70%. The priority is calculated based on the track data type and user priority weight.

3. The film review system supporting multi-track customized playback and version switching according to claim 1, characterized in that: The multi-track management module includes a multi-track data acquisition unit and a multi-track data storage unit; The multi-track data acquisition unit performs multi-track data acquisition and processing of film and television content through film and television editing software interfaces and API integration tools, and generates original multi-track datasets; The multi-track data storage unit partitions and stores the original multi-track dataset based on a distributed database system, and establishes a track index table to generate and optimize multi-track data. The original multi-track dataset includes video frame rate parameters, audio sampling rate parameters, subtitle timestamp parameters, and special effects layer parameters. The distributed database system adopts a Hadoop and Spark architecture. The track index table includes track type identifiers, timeline start positions, and content feature hash values.

4. A film review system supporting multi-track customized playback and version switching according to claim 1, characterized in that: The playback customization module includes a user preference analysis unit and a playback parameter adjustment unit; The user preference analysis unit performs preference feature extraction processing based on user input preference data and a collaborative filtering algorithm to generate a user preference feature vector. The playback parameter adjustment unit dynamically adjusts the stored optimized multi-track data based on the user preference feature vector, including playback speed parameters, volume balance parameters, and subtitle synchronization parameters, and generates a customized playback sequence. The user input preference data includes playback speed preference value, volume preference level, and subtitle language preference. The collaborative filtering algorithm constructs a preference matrix based on the user's historical behavior data, and the customized playback sequence is represented as a combination of multi-track parameters aligned with the time axis.

5. A film review system supporting multi-track customized playback and version switching according to claim 4, characterized in that: The playback parameter adjustment unit performs dynamic parameter adjustment operations as follows: S61. Constructing a Preference Matrix: Based on historical user behavior data, including playback speed preference values, volume preference levels, and subtitle language preferences, construct a user-item preference matrix. The rows represent user identifiers, and the columns represent playback parameter types, including playback speed, volume balance, and subtitle synchronization. (Matrix elements) Indicates user For parameters Preference rating; For missing values, the average user preference is used to fill them in, and the singular value decomposition (SVD) is used to reduce the dimensionality of the preference matrix. in For the user-item preference matrix, For the user feature matrix, It is a singular value matrix. For the project feature matrix, This is the transpose of the project feature matrix; Extracting user feature vectors As input to the user preference feature vector; S62. Obtain user preference feature vectors and store optimized multi-track data; S63. Based on a deep learning model, perform track coordination analysis on storage-optimized multi-track data to generate track coordination feature data; S64. Calculate the playback parameter adjustment weights based on the user preference feature vector and track coordination feature data, and generate the parameter adjustment weight matrix; S65. An interpolation algorithm is used to adjust the playback speed parameters, volume balance parameters, and subtitle synchronization parameters of multi-track data in real time according to the parameter adjustment weight matrix, and output a customized playback sequence. The deep learning model is a convolutional neural network (CNN) and a long short-term memory network (LSTM). The track collaborative feature data includes video-audio synchronization coefficients and subtitle-effect alignment parameters. The interpolation algorithm is linear interpolation and spline interpolation.

6. A film review system supporting multi-track customized playback and version switching according to claim 1, characterized in that: The user interaction module includes an input receiving unit and a feedback output unit; The input receiving unit receives user input data through a graphical user interface (GUI) and a voice recognition interface, including playback customization preference data and version switching request data. The feedback output unit renders the customized playback sequence and target version switching instructions into a real-time visualization interface and outputs it to the display device. The graphical user interface supports drag-and-drop track editing, and the real-time visualization interface includes multi-track waveforms and a version switching progress bar.

Images