A blockchain-based movie and television content storage management method and system

Abstract

The application discloses a kind of based on block chain's film and television content storage management method and system, it is related to image data storage technical field, the method includes: based on the video sequence and audio stream data of film and television content, by extracting scene conversion node, film and television content is segmented into several content segments, generates content semantic weight map, and records the semantic feature set of each content segment;According to content semantic weight map, the hierarchical storage strategy of each content segment is established, and the position index and access permission information of each content segment are recorded by block chain accounting mechanism;According to content pre-fetching priority queue, the storage level allocation of each content segment in hierarchical storage strategy is adjusted, and the migration and cache update of each content segment between different storage nodes are completed based on position index and access permission information.The present application has the advantages of semantic perception intelligent hierarchical storage and dynamic pre-fetching scheduling.

Description

Technical Field

[0001] This invention relates to the field of image data storage technology, and more specifically, to a blockchain-based method and system for managing film and television content storage. Background Technology

[0002] With the rapid development of the digital entertainment industry and the ever-growing demand for high-quality film and television content, the amount of data in film and television content is experiencing explosive growth. Modern film and television platforms need to process massive amounts of video data in their server systems, including films and television works of different resolutions, formats, and durations. This content not only consumes a large amount of storage space but also needs to support the concurrent access needs of hundreds of millions of users worldwide. While traditional content delivery networks can alleviate bandwidth pressure to some extent, existing server-side storage management methods are no longer sufficient to meet the needs of content management and resource allocation in the face of diversified user viewing habits and fragmented content consumption patterns. Therefore, a more intelligent storage management solution is urgently needed.

[0003] Currently, the storage management of film and television content mainly adopts a hierarchical storage strategy based on access frequency and content distribution network technology. Popular content is stored in a high-speed cache layer by statistically analyzing user access data, while less popular content is migrated to low-cost storage devices. Some manufacturers in the industry have adopted content fragmentation, which divides complete film and television files into multiple segments for distributed storage to improve the system's concurrent processing capabilities and availability. Other manufacturers choose to use content deduplication and compression algorithms to optimize storage space utilization and improve content access response speed.

[0004] However, existing storage management technologies for film and television content on servers still have significant shortcomings: traditional hierarchical storage strategies based on access frequency often ignore the semantic value differences of the film and television content itself, and cannot identify the actual importance of content segments, resulting in important segments being misclassified as low-value, unpopular content; secondly, existing prediction algorithms mainly rely on statistical methods and lack a deep understanding of the semantic features of film and television content, making it impossible to accurately predict users' viewing needs for specific content segments. This not only affects the accuracy of storage strategies but also reduces the resource utilization efficiency of server systems.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] To address the problems in related technologies, this invention proposes a blockchain-based method and system for managing film and television content storage. This system has the advantages of semantically aware intelligent hierarchical storage and dynamic prefetching scheduling, thereby solving the problems in existing technologies where server-side storage management lacks the ability to identify the semantic value of film and television content, resulting in slow access response for important content segments and unreasonable allocation of storage resources.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows:

[0008] According to one aspect of the present invention, a blockchain-based method for managing film and television content storage is provided. This method includes: based on video sequence and audio stream data of the film and television content, by extracting scene transition nodes, dividing the film and television content into several content segments, generating a content semantic weight map, and recording the semantic feature set of each content segment; establishing a hierarchical storage strategy for each content segment based on the content semantic weight map, and recording the location index and access permission information of each content segment through a blockchain ledger mechanism; constructing a content prefetching priority queue based on the semantic feature set of each content segment and user viewing behavior data, adjusting the storage level allocation of each content segment in the hierarchical storage strategy according to the content prefetching priority queue, and completing the migration and cache update of each content segment between different storage nodes based on the location index and access permission information.

[0009] Furthermore, segmenting the film and television content into several content segments by extracting scene transition nodes includes: extracting video frame sequences and audio features from the video sequence and audio stream data of the film and television content; detecting the visual difference between consecutive frames in the video frame sequence and setting a visual difference threshold; marking a candidate scene transition node when the visual difference between consecutive frames exceeds the visual difference threshold; extracting the audio energy change rate and silent segment distribution from the audio features; confirming the candidate scene transition node as a valid scene transition node when the audio energy change rate corresponding to the candidate scene transition node exceeds a preset energy change threshold, or when a silent segment with a duration exceeding a preset duration is detected; and segmenting the film and television content into several content segments based on the confirmed valid scene transition nodes.

[0010] Furthermore, detecting the visual difference between consecutive frames in a video frame sequence includes: extracting the RGB color histograms and edge gradient features of adjacent video frames in the video frame sequence; calculating the Bach distance of the RGB color histograms and the Euclidean distance of the edge gradient features between adjacent frames; constructing a two-dimensional feature vector based on the Bach distance and the Euclidean distance; obtaining the visual similarity by calculating the cosine of the angle between the two-dimensional feature vector and the preset reference vector; and using the reciprocal of the visual similarity as the visual difference.

[0011] Furthermore, extracting the distribution of silent segments in audio features includes: performing a short-time Fourier transform on the audio signal to obtain the spectral energy distribution; when the average spectral energy within a continuous time window is lower than a preset background noise energy threshold and the duration exceeds a preset silence duration, it is marked as a silent segment; extracting the audio energy change rate in audio features includes: extracting the effective values ​​of the audio signal in adjacent time windows, and calculating the ratio of the difference between the effective values ​​of the audio signal in the preceding and following time windows to the effective value of the audio signal in the previous time window to obtain the audio energy change rate.

[0012] Furthermore, generating the content semantic weight map includes: extracting the audio signal amplitude values ​​of each content segment, calculating the ratio of the standard deviation to the average value of the audio signal amplitude values ​​within each content segment as the sentiment fluctuation index; obtaining the user viewing time of each content segment from the server-side user behavior logs, and calculating the ratio of the user viewing time to the total duration of the content segment as the attention coefficient; multiplying the sentiment fluctuation index and the attention coefficient of each content segment to obtain the semantic importance weight value of each content segment; establishing a mapping coordinate system with the start time of each content segment in the film and television content as the horizontal axis and the corresponding semantic importance weight value as the vertical axis; and supplementing intermediate weight values ​​between adjacent content segments using a linear interpolation method based on the mapping coordinate system to form a continuous weight curve, thus obtaining the content semantic weight map.

[0013] Furthermore, the hierarchical storage strategy for each content segment based on the content semantic weight map includes: calculating the average and variance of the semantic importance weight values ​​in the content semantic weight map, using the average plus or minus one time the variance as the hierarchical threshold boundary; classifying content segments with semantic importance weight values ​​greater than the average plus or minus one time the variance into the first storage level; classifying content segments with semantic importance weight values ​​between the average minus one time the variance and the average plus one time the variance into the second storage level; classifying content segments with semantic importance weight values ​​less than the average minus one time the variance into the third storage level; and allocating the corresponding storage device type according to the storage level into which each content segment is classified, in order to establish the hierarchical storage strategy.

[0014] Furthermore, recording the location index and access permission information of each content fragment through the blockchain ledger mechanism includes: generating a unique identifier for each content fragment, and combining the start time, storage level, and unique identifier of each content fragment to obtain a location index; setting access permission levels as access permission information based on the semantic feature set of each content fragment; organizing the location index and access permission information of each content fragment into data blocks according to a time series; sorting the time series according to the start time of the content fragments in chronological order; and associating adjacent data blocks through a hash linking algorithm to form a chain record structure, thereby realizing the blockchain storage of the location index and access permission information of each content fragment.

[0015] Furthermore, based on the semantic feature set of each content fragment, setting access permission levels as access permission information includes: extracting the sentiment fluctuation index and attention coefficient from the semantic feature set of each content fragment; calculating the weighted average of the sentiment fluctuation index and attention coefficient as the content sensitivity index; setting access permission thresholds according to the content sensitivity index, setting content fragments that exceed the access permission threshold as restricted access level, and setting content fragments that do not exceed the access permission threshold as open access level, so as to determine the access permission information of each content fragment.

[0016] Furthermore, constructing a content prefetching priority queue based on the semantic feature set of each content segment and user viewing behavior data includes: extracting the access frequency and jump behavior identifier of each content segment from the user viewing behavior data; the jump behavior identifier records whether the user skipped the content segment; initially sorting each content segment according to its semantic importance weight value, and adjusting the sorting position of content segments with jump behavior identifiers to the back by a preset position; identifying content segments with an access frequency higher than a preset frequency threshold as hot segments, and advancing the sorting position of hot segments by a preset position; according to the adjusted sorting position, assigning the first half of the content segments to the high-priority prefetching queue and the second half of the content segments to the low-priority prefetching queue; and using the hierarchical structure of the high-priority prefetching queue and the low-priority prefetching queue as the content prefetching priority queue.

[0017] According to another aspect of the present invention, a blockchain-based film and television content storage management system is also provided. This system includes: a content analysis module, used to segment the film and television content into several content segments based on video sequence and audio stream data, by extracting scene transition nodes, generating a content semantic weight map, and recording the semantic feature set of each content segment; a storage management module, used to establish a hierarchical storage strategy for each content segment based on the content semantic weight map, and record the location index and access permission information of each content segment through a blockchain ledger mechanism; and a dynamic scheduling module, used to construct a content prefetching priority queue based on the semantic feature set of each content segment and user viewing behavior data, adjust the storage level allocation of each content segment in the hierarchical storage strategy according to the content prefetching priority queue, and complete the migration and cache update of each content segment between different storage nodes based on the location index and access permission information.

[0018] The beneficial effects of this invention are as follows:

[0019] (1) This invention segments content by extracting scene transition nodes from video sequences and audio stream data of film and television content, and calculates the emotional fluctuation index based on the amplitude value of the audio signal and generates a content semantic weight map based on the user's viewing time. This enables the server to automatically identify important content such as climax scenes and wonderful dialogues in film and television works and store them in the high-speed cache layer first. This breaks through the technical bottleneck of the traditional server-side hierarchical storage strategy based on access frequency that ignores the differences in the semantic value of content, and helps to improve the data transmission speed and playback smoothness of the server when responding to users accessing popular content.

[0020] (2) This invention establishes a three-level storage strategy for servers by calculating the average value and variance of semantic importance weight values ​​to set the grading threshold boundary. At the same time, it uses the blockchain ledger mechanism to record the location index and access permission information of each content fragment in the server storage node. This solves the technical problem that existing server-side prediction algorithms lack a deep understanding of the semantic features of film and television content. Compared with the traditional unified server storage mode, it can save more than 30% of storage costs and avoid the situation where important fragments are incorrectly classified as low-value unpopular content in server storage management.

[0021] (3) Based on the semantic feature set of each content segment and the user jump behavior identifier, the present invention constructs a hierarchical content prefetch priority queue. By identifying hot segments and adjusting the sorting position, the dynamic allocation of server storage levels is realized. This enables the server to predict the content segments that users may watch in advance and perform cache preloading, which effectively improves the efficiency of server system resource utilization and reduces the waiting time after the user clicks to play. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a blockchain-based film and television content storage management method according to an embodiment of the present invention.

[0024] Figure 2 This is a specific implementation diagram of a blockchain-based film and television content storage management method according to an embodiment of the present invention, which divides film and television content into several content segments;

[0025] Figure 3 This is a schematic diagram of a blockchain-based film and television content storage and management system according to an embodiment of the present invention. Detailed Implementation

[0026] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0027] According to an embodiment of the present invention, a blockchain-based method and system for managing film and television content storage is provided.

[0028] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a blockchain-based method for managing film and television content storage is provided, the method comprising:

[0029] S1. Based on video sequence and audio stream data of film and television content, the film and television content is divided into several content segments by extracting scene transition nodes, generating a content semantic weight map, and recording the semantic feature set of each content segment;

[0030] S2. Establish a hierarchical storage strategy for each content segment based on the content semantic weight map, and record the location index and access permission information of each content segment through the blockchain ledger mechanism;

[0031] S3. Construct a content prefetch priority queue based on the semantic feature set of each content fragment and user viewing behavior data. Adjust the storage level allocation of each content fragment in the hierarchical storage strategy according to the content prefetch priority queue, and complete the migration and cache update of each content fragment between different storage nodes based on the location index and access permission information.

[0032] Specifically, the semantic feature set of each content segment includes: emotional fluctuation index, attention coefficient, semantic importance weight value, visual difference, audio energy change rate, silent segment distribution, scene transition nodes, and start time in film and television content.

[0033] In one embodiment, S1, segmenting the film and television content into several content segments by extracting scene transition nodes, includes: extracting video frame sequences and audio features from the video sequence and audio stream data of the film and television content; detecting the visual difference between consecutive frames in the video frame sequence and setting a visual difference threshold; marking a candidate scene transition node when the visual difference between consecutive frames exceeds the visual difference threshold; extracting the audio energy change rate and silent segment distribution from the audio features; confirming the candidate scene transition node as a valid scene transition node when the audio energy change rate corresponding to the candidate scene transition node exceeds a preset energy change threshold, or when a silent segment with a duration exceeding a preset duration is detected; and segmenting the film and television content into several content segments based on the confirmed valid scene transition nodes.

[0034] In one embodiment, detecting the visual difference between consecutive frames in a video frame sequence includes: extracting the RGB color histograms and edge gradient features of adjacent video frames in the video frame sequence; calculating the Bach distance between the RGB color histograms and the Euclidean distance between the edge gradient features of adjacent frames; constructing a two-dimensional feature vector based on the Bach distance and the Euclidean distance; obtaining the visual similarity by calculating the cosine of the angle between the two-dimensional feature vector and a preset reference vector; and using the reciprocal of the visual similarity as the visual difference.

[0035] In one embodiment, extracting the distribution of silent segments in audio features includes: performing a short-time Fourier transform on the audio signal to obtain the spectral energy distribution; when the average spectral energy within a continuous time window is lower than a preset background noise energy threshold and the duration exceeds a preset silence duration, it is marked as a silent segment; extracting the rate of change of audio energy in audio features includes: extracting the effective values ​​of audio signals in adjacent time windows, and calculating the ratio of the difference between the effective values ​​of audio signals in the preceding and following time windows to the effective value of audio signals in the previous time window to obtain the rate of change of audio energy.

[0036] Specifically, taking the content segmentation of a 120-minute film stored on a video platform server as an example, according to the storage management method proposed in this invention, the server first extracts a video frame sequence of 25 frames per second from the stored H.264 encoded video file, and at the same time extracts audio stream data with a sampling rate of 44.1kHz from the AAC encoded audio track. The video frame sequence is numbered from frame 1 to frame 180000 in chronological order, and the audio stream data is divided into a total of 7200 windows with a duration of 1 second for analysis and processing.

[0037] Specifically, when detecting the visual differences between consecutive frames in the movie video frame sequence, the server constructs a 256-level grayscale histogram for each of the RGB three color channels of the video frame. By counting the number of pixels at each grayscale level, three 256-dimensional color distribution vectors are obtained. At the same time, the Sobel operator is used to perform edge detection on each video frame to calculate the gradient magnitude in the horizontal and vertical directions. Pixels with gradient magnitudes greater than a preset threshold of 50 are marked as edge pixels. The distribution of edge pixels is statistically analyzed to form an edge gradient feature vector.

[0038] It should be noted that the Sobel operator is a commonly used edge detection operator that detects image edges by calculating an approximate value of the gradient of the image brightness function. This operator uses two 3×3 convolution kernels to calculate the gradients in the horizontal and vertical directions, respectively. After convolving these two kernels with the image, the gradient magnitude is obtained by calculating the square root of the sum of the squares of the gradients in the two directions. In this invention, the Sobel operator is used to extract the edge gradient features of video frames, thereby providing the edge information basis for subsequent calculation of the visual differences between adjacent frames.

[0039] Specifically, when calculating the visual difference between adjacent frames, the server denotes the RGB color histograms of the previous and current frames as H1 and H2, respectively, and calculates the Barthold distance BC(H1(i), H2(i)), where i represents the i-th component of the histogram; simultaneously, it calculates the Euclidean distance ED(E1(i), E2(i)) between the edge gradient feature vectors E1 and E2 of the two frames, and combines the Barthold distance and Euclidean distance after normalization to form a two-dimensional feature vector [BC]. n ED n The cosine of the angle between this vector and the preset reference vector [0.8, 0.6] is calculated as cos(θ) = ([BC...). n ED n ]×[0.8,0.6]) / (||[BC n ED n The visual similarity is obtained by ]||×||[0.8,0.6]||, and 1 / cos(θ) is used as the visual difference value.

[0040] Specifically, when the calculated visual difference value exceeds the preset visual difference threshold of 1.5, the server marks the frame as a candidate scene transition node and records its timestamp. When extracting the distribution of silent segments in audio features, the server uses a short-time Fourier transform on the audio signal corresponding to the timestamp of the candidate scene transition node to transform the audio signal from the time domain to the frequency domain and sets a 50% overlapping Hamming window. The spectral energy distribution is obtained by calculating the power spectral density of the preset frequency range within each time-frequency window. When the average spectral energy within 5 consecutive time windows is lower than the preset background noise energy threshold and the duration exceeds the preset silence duration of 0.5 seconds, the time period is marked as a silent segment and its start and end times are recorded.

[0041] Specifically, when extracting the audio energy change rate from the audio features, the server extracts the effective values ​​of the audio signals in the adjacent time windows before and after the candidate scene transition node. The effective value RMS is obtained by calculating the square root of the mean of the square of the audio signal amplitude. Let the effective value of the audio signal in the previous time window be RMS1 and the effective value of the audio signal in the next time window be RMS2. Then the formula for calculating the audio energy change rate is (RMS2-RMS1) / RMS1×100%. When the calculated audio energy change rate exceeds the preset energy change threshold of 30%, or when a silent segment with a duration of more than 0.5 seconds is detected, the candidate scene transition node is confirmed as a valid scene transition node.

[0042] In one embodiment, generating a content semantic weight map in S1 includes: extracting the audio signal amplitude values ​​of each content segment, calculating the ratio of the standard deviation to the average value of the audio signal amplitude values ​​within each content segment as an emotional fluctuation index; obtaining the user viewing time of each content segment from the server-side user behavior logs, and calculating the ratio of the user viewing time to the total duration of the content segment as an attention coefficient; multiplying the emotional fluctuation index and the attention coefficient of each content segment to obtain the semantic importance weight value of each content segment; establishing a mapping coordinate system with the start time of each content segment in the film and television content as the horizontal axis and the corresponding semantic importance weight value as the vertical axis; and supplementing intermediate weight values ​​between adjacent content segments using a linear interpolation method based on the mapping coordinate system to form a continuous weight curve, thereby obtaining a content semantic weight map.

[0043] Specifically, after the server divides the film and television content into several content segments based on the confirmed valid scene transition nodes, it begins to generate a content semantic weight map. In the above embodiment, the 120-minute film and television work is divided into 86 content segments. The server extracts the audio signal amplitude value of each content segment, obtains the amplitude sequence by calculating the absolute value of the audio signal at each sampling point, and then calculates the standard deviation σ and the average value μ of the audio signal amplitude value in each content segment. The ratio of the standard deviation to the average value σ / μ is used as the emotional fluctuation index of the content segment.

[0044] Specifically, the server queries the user behavior log database for statistical data on user viewing time for each content segment of the film / TV show. It then calculates the average viewing time by aggregating the actual viewing time of all users for that content segment over the past 30 days, and uses the ratio of user viewing time to the total duration of the content segment as the attention coefficient. The server multiplies the emotional fluctuation index of each content segment with the attention coefficient to obtain the semantic importance weight value for each content segment. A mapping coordinate system is established with the start time of each content segment in the film / TV show as the horizontal axis and the corresponding semantic importance weight value as the vertical axis.

[0045] Specifically, based on the established mapped coordinate system, the server uses linear interpolation to supplement intermediate weight values ​​between adjacent content segments, forming a continuous weight curve. The server performs interpolation calculations at preset time intervals within the time interval between every two adjacent content segments. It calculates the weight value at intermediate moments point by point using the linear interpolation formula y=y1+(y2-y1)×(x-x1) / (x2-x1), where y1 and y2 are the semantic importance weight values ​​of the preceding and following content segments, x1 and x2 are the start times of the preceding and following content segments, x is the intermediate moment for which weight values ​​need to be calculated, and y is the weight value corresponding to that intermediate moment, ensuring the continuity of the weight curve across the entire time axis. Finally, a content semantic weight map composed of the mapped coordinate system and the continuous weight curve is obtained. This content semantic weight map visually displays the semantic importance distribution of film and television content at various moments across the entire time axis in a two-dimensional graphical form, thus providing data support for subsequent hierarchical storage strategies.

[0046] In one embodiment, S2, establishing a hierarchical storage strategy for each content segment based on the content semantic weight map, includes: calculating the average and variance of the semantic importance weight values ​​in the content semantic weight map, using the average plus or minus one time the variance as the hierarchical threshold boundary; classifying content segments with semantic importance weight values ​​greater than the average plus or minus one time the variance into the first storage level; classifying content segments with semantic importance weight values ​​between the average minus one time the variance and the average plus one time the variance into the second storage level; classifying content segments with semantic importance weight values ​​less than the average minus one time the variance into the third storage level; and allocating the corresponding storage device type according to the storage level to which each content segment is classified, in order to establish a hierarchical storage strategy.

[0047] Specifically, the storage device types include a high-speed cache layer (allocation storage first storage level), a standard storage layer (allocation storage second storage level), and low-cost storage devices (allocation storage third storage level). This invention, through this hierarchical storage strategy, allocates content fragments with higher semantic importance to faster storage devices, thus ensuring rapid access to important content and meeting the practical needs of film and television content storage management.

[0048] Specifically, in the above embodiment, the server calculates the average and variance of the semantic importance weight values ​​of 86 content segments in the semantic weight map of the 120-minute film. The average value μ is calculated to be 0.85 and the variance σ is obtained by traversing all weight value sequences. 2 With a mean of 0.16 and a standard deviation σ of 0.4, the upper threshold is 1.25 and the lower threshold is 0.45, defined by the mean plus or minus one standard deviation. Based on this, the server assigns content segments with semantic importance weights greater than 1.25 to the first storage level, content segments with weights between 0.45 and 1.25 to the second storage level, and content segments with weights less than 0.45 to the third storage level. These are then allocated to the cache layer, standard storage layer, and low-cost storage devices, respectively, thus establishing a tiered storage strategy.

[0049] In one embodiment, S2, recording the location index and access permission information of each content fragment through a blockchain ledger mechanism, includes: generating a unique identifier for each content fragment and combining the start time, storage level, and unique identifier of each content fragment to obtain a location index; setting an access permission level as access permission information based on the semantic feature set of each content fragment; organizing the location index and access permission information of each content fragment into data blocks according to a time series; sorting the time series according to the start time of the content fragments in chronological order; and associating adjacent data blocks through a hash linking algorithm to form a chain record structure, thereby realizing the blockchain storage of the location index and access permission information of each content fragment.

[0050] In one embodiment, setting access permission levels as access permission information based on the semantic feature set of each content fragment includes: extracting the sentiment fluctuation index and attention coefficient from the semantic feature set of each content fragment; calculating the weighted average of the sentiment fluctuation index and attention coefficient as a content sensitivity index; setting access permission thresholds according to the content sensitivity index, setting content fragments that exceed the access permission threshold as restricted access level, and setting content fragments that do not exceed the access permission threshold as open access level, so as to determine the access permission information of each content fragment.

[0051] Specifically, the server generates a unique identifier for each content fragment. A 32-bit hexadecimal string is generated using the UUID algorithm as the unique identifier. The start time of each content fragment is converted to a timestamp format, and the storage level number and unique identifier are concatenated to obtain the location index. The server extracts the sentiment fluctuation index and attention coefficient from the semantic feature set of each content fragment, and calculates their weighted average as the content sensitivity index. In the above embodiment, the sentiment fluctuation index is calculated with a weight of 0.6, and the attention coefficient with a weight of 0.4. Based on a preset content sensitivity threshold of 0.8, content fragments with a sensitivity index exceeding 0.8 are set to a restricted access level, while those below 0.8 are set to an open access level, forming access permission information for each content fragment. The location index and access permission information are organized into data blocks according to the chronological order of the content fragment's start time. The hash value of each data block is calculated using the SHA-256 hash algorithm and linked with the hash value of the previous data block to form a blockchain storage structure.

[0052] In one embodiment, S3, constructing a content prefetching priority queue based on the semantic feature set of each content segment and user viewing behavior data, includes: extracting the access frequency and jump behavior identifier of each content segment from the user viewing behavior data; the jump behavior identifier records whether the user skipped the content segment; initially sorting each content segment according to its semantic importance weight value, and adjusting the sorting position of content segments with jump behavior identifiers to the back by a preset position; identifying content segments with access frequencies higher than a preset frequency threshold as hot segments, and advancing the sorting position of hot segments by a preset position; according to the adjusted sorting position, assigning the first half of the content segments to the high-priority prefetching queue and the second half of the content segments to the low-priority prefetching queue; and using the hierarchical structure of the high-priority prefetching queue and the low-priority prefetching queue as the content prefetching priority queue.

[0053] Specifically, in the above embodiment, the server collects user access data for each content segment of the 120-minute film / TV show from the user behavior log database over the past 7 days. It calculates the access frequency by analyzing the number of clicks on 86 content segments and identifies user skipping behavior markers. The 15th content segment has an access frequency of 1200 times per day. By detecting user skipping of the playback progress bar, it is found that 15% of this segment is skipped, thus marking it as having skipping behavior. The server initially sorts each content segment according to its semantic importance weight. Content segments with skipping behavior markers are moved 10 positions backward. Content segments with access frequencies exceeding a preset threshold of 800 times per day are identified as hot segments, and their positions are moved forward 5 positions. Based on the adjusted sort positions, the first 43 content segments are placed in a high-priority prefetch queue, and the last 43 content segments are placed in a low-priority prefetch queue, forming a hierarchical content prefetch priority queue.

[0054] Specifically, based on the sorting results of the content prefetch priority queue, the server automatically adjusts the storage level allocation of each content fragment in the hierarchical storage strategy. Content fragments in the high-priority prefetch queue are prioritized one storage level higher (i.e., high-priority content fragments originally located in the standard storage layer are migrated to the cache layer, and high-priority content fragments originally located in low-cost storage devices are migrated to the standard storage layer), while content fragments in the low-priority prefetch queue are downgraded one storage level. Based on the location index and access permission information recorded in the blockchain, the server queries the specific location of each content fragment in different storage nodes, completes the data migration of content fragments between storage nodes through a high-speed intranet transmission channel, and updates the content cache in the cache layer in real time. This ensures that users can obtain the best response speed and playback experience when accessing popular content, thereby achieving intelligent film and television content storage management based on semantic understanding and user behavior prediction.

[0055] like Figure 3 As shown, according to another embodiment of the present invention, a blockchain-based film and television content storage and management system is also provided, which includes:

[0056] Content analysis module 1 is used to extract scene transition nodes from video sequences and audio stream data of film and television content, divide film and television content into several content segments, generate a content semantic weight map, and record the semantic feature set of each content segment.

[0057] Storage management module 2 is used to establish a hierarchical storage strategy for each content fragment based on the content semantic weight map, and to record the location index and access permission information of each content fragment through a blockchain ledger mechanism;

[0058] The dynamic scheduling module 3 is used to construct a content prefetch priority queue based on the semantic feature set of each content fragment and user viewing behavior data, adjust the storage level allocation of each content fragment in the hierarchical storage strategy according to the content prefetch priority queue, and complete the migration and cache update of each content fragment between different storage nodes based on the location index and access permission information.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A blockchain-based method for storing and managing film and television content, characterized in that, include: Based on video sequences and audio stream data of film and television content, the film and television content is divided into several content segments by extracting scene transition nodes, generating a content semantic weight map, and recording the semantic feature set of each content segment; A hierarchical storage strategy for each content fragment is established based on the content semantic weight map, and the location index and access permission information of each content fragment are recorded through a blockchain ledger mechanism. A content prefetch priority queue is constructed based on the semantic feature set of each content fragment and user viewing behavior data. The storage level allocation of each content fragment in the hierarchical storage strategy is adjusted according to the content prefetch priority queue. The migration and cache update of each content fragment between different storage nodes are completed based on the location index and access permission information.

2. The method for storing and managing film and television content based on blockchain according to claim 1, characterized in that, The step of segmenting film and television content into several content fragments by extracting scene transition nodes includes: Extract video frame sequences and audio features from the video sequence and audio stream data of the film and television content; Detect the visual difference between consecutive frames in a video frame sequence and set a visual difference threshold; When the visual difference between consecutive frames exceeds the visual difference threshold, it is marked as a candidate scene transition node; Extract the rate of change of audio energy and the distribution of silent segments from audio features; When the audio energy change rate corresponding to a candidate scene transition node exceeds a preset energy change threshold, or when a silent segment with a duration exceeding a preset duration is detected, the candidate scene transition node is confirmed as a valid scene transition node. Based on the confirmed valid scene transition nodes, the film and television content is divided into several content segments.

3. The method for storing and managing film and television content based on blockchain according to claim 2, characterized in that, The visual differences between consecutive frames in the detected video frame sequence include: Extract RGB color histograms and edge gradient features from adjacent video frames in a video frame sequence; Calculate the Bach distance of the RGB color histograms and the Euclidean distance of the edge gradient features between adjacent frames; Two-dimensional feature vectors are constructed based on Bach's distance and Euclidean distance. Visual similarity is obtained by calculating the cosine of the angle between the two-dimensional feature vector and the preset reference vector, and the reciprocal of the visual similarity is used as the visual difference.

4. The method for storing and managing film and television content based on blockchain according to claim 2, characterized in that, The extraction of silent segment distribution in audio features includes: performing a short-time Fourier transform on the audio signal to obtain the spectral energy distribution; when the mean spectral energy within a continuous time window is lower than a preset background noise energy threshold and the duration exceeds a preset silence duration, it is marked as a silent segment. The extraction of audio energy change rate from audio features includes: extracting the effective values ​​of audio signals from adjacent time windows, and calculating the ratio of the difference between the effective values ​​of audio signals in the preceding and following time windows to the effective value of audio signals in the previous time window, to obtain the audio energy change rate.

5. The method for storing and managing film and television content based on blockchain according to claim 1, characterized in that, The generated content semantic weight map includes: Extract the audio signal amplitude values ​​of each content segment, and calculate the ratio of the standard deviation to the mean of the audio signal amplitude values ​​within each content segment as the emotional fluctuation index; The user viewing time for each content segment is obtained from the server-side user behavior logs, and the ratio of the user viewing time to the total duration of the content segment is calculated as the attention coefficient. The semantic importance weight value of each content segment is obtained by multiplying the emotional fluctuation index of each content segment with the attention coefficient. A mapping coordinate system is established with the start time of each content segment in the film and television content as the horizontal axis and the corresponding semantic importance weight value as the vertical axis. Based on the mapped coordinate system, intermediate weight values ​​are added between adjacent content segments using linear interpolation to form a continuous weight curve, thus obtaining a content semantic weight map.

6. The method for storing and managing film and television content based on blockchain according to claim 1, characterized in that, The hierarchical storage strategy for each content fragment based on the content semantic weight map includes: Calculate the average and variance of the semantic importance weight values ​​in the content semantic weight map, and use the average plus or minus one time the variance as the classification threshold boundary; Content fragments with semantic importance weights greater than the average plus one variance are classified into the first storage level; Content segments with semantic importance weights between the average minus one variance and the average plus one variance are classified into the second storage level; Content fragments with semantic importance weights less than the average minus one time the variance are classified into the third storage level; Based on the storage level to which each content fragment belongs, the corresponding storage device type is allocated to establish a tiered storage strategy.

7. A blockchain-based film and television content storage management method according to claim 6, characterized in that, The method of recording the location index and access permission information of each content fragment through the blockchain ledger mechanism includes: Generate a unique identifier for each content fragment, and combine the start time, storage level, and unique identifier of each content fragment to obtain a location index; Based on the semantic feature set of each content fragment, an access permission level is set as access permission information; The location index and access permission information of each content segment are organized into data blocks according to the time series; the time series is determined by sorting the content segments in chronological order based on their start time. Adjacent data blocks are linked by a hash chaining algorithm to form a chain-like record structure, enabling blockchain storage of location indexes and access permission information for each content fragment.

8. The method for storing and managing film and television content based on blockchain according to claim 7, characterized in that, The access permission level, set based on the semantic feature set of each content fragment, includes the following: Extract the sentiment fluctuation index and attention coefficient from the semantic feature set of each content fragment; The weighted average of the emotional fluctuation index and the attention coefficient is calculated as a content sensitivity indicator; Access permission thresholds are set based on content sensitivity indicators. Content segments that exceed the access permission threshold are set to restricted access level, while content segments that do not exceed the access permission threshold are set to open access level, in order to determine the access permission information for each content segment.

9. A blockchain-based film and television content storage management method according to claim 1, characterized in that, The construction of the content prefetching priority queue based on the semantic feature set of each content fragment and user viewing behavior data includes: Extract the access frequency and jump behavior identifiers of each content segment from user viewing behavior data; the jump behavior identifiers record whether the user skipped the content segment; The content segments are initially sorted according to their semantic importance weight values, and the sorting positions of the content segments with jump behavior indicators are adjusted to the next preset position. Content segments with access frequency higher than a preset frequency threshold are identified as hot segments, and their sorting position is moved up a preset number of positions. Based on the adjusted sorting position, the first half of the content fragments are placed into the high-priority prefetch queue, and the second half of the content fragments are placed into the low-priority prefetch queue. A hierarchical structure of high-priority prefetch queues and low-priority prefetch queues is used as the content prefetch priority queue.

10. A blockchain-based film and television content storage management system, used to implement the blockchain-based film and television content storage management method according to any one of claims 1-9, characterized in that, The system includes: The content analysis module is used to extract scene transition nodes from video sequences and audio stream data of film and television content, divide the film and television content into several content segments, generate a content semantic weight map, and record the semantic feature set of each content segment. The storage management module is used to establish a hierarchical storage strategy for each content fragment based on the content semantic weight map, and to record the location index and access permission information of each content fragment through a blockchain ledger mechanism. The dynamic scheduling module is used to construct a content prefetch priority queue based on the semantic feature set of each content fragment and user viewing behavior data, adjust the storage level allocation of each content fragment in the hierarchical storage strategy according to the content prefetch priority queue, and complete the migration and cache update of each content fragment between different storage nodes based on location index and access permission information.

Images