A processing method and system for short video playing

Abstract

The application relates to the technical field of video playing, and discloses a processing method and system for short video playing, which comprises the following steps: acquiring user information and traversing in a historical user library, determining a multiplex playing strategy or a model playing strategy according to a traversal result, acquiring a sample interaction data set and constructing a strategy learning model according to the sample interaction data set, analyzing target playing environment data to generate a playing adaptation mark, analyzing all playing environment sub-data of the same type based on the playing adaptation mark and constructing a playing environment influence sequence, determining an association result of a playing mixed influence set according to an association rule algorithm, substituting data not associated in the playing mixed influence set and the association result into a Bayesian model, determining a playing improvement strategy, and judging whether the playing improvement strategy needs to be adjusted based on the relationship between playing resource energy consumption and improved resource energy consumption. The application ensures the flexibility and environmental adaptability of short video playing.

Description

Technical Field

[0001] This invention relates to the field of video playback technology, and more specifically, to a processing method and system for short video playback. Background Technology

[0002] Driven by the digital wave, short videos, with their short duration, rich content, and rapid dissemination, have become a core carrier of information dissemination and mass entertainment. However, current short video playback has certain limitations in terms of environmental adaptability and matching user habits. Different users have different operating habits, while existing short video playback usually uses a fixed adjustment method, which cannot capture changes in user habits. For example, some users are used to double-tapping to pause, while others are used to single-tapping. The short video playback and user behavior habits are misaligned. In terms of environmental adaptation, different scenarios have differences in both light intensity and noise levels. In bright outdoor environments, the screen is prone to glare, causing loss of image details, and users need to manually increase the brightness to view clearly. In dim nighttime environments, excessively high video brightness can cause visual stimulation and eye discomfort. In terms of sound, noisy public places require users to repeatedly turn up the volume to hear the audio content, while in quiet environments such as libraries and conference rooms, sudden video sounds can interfere with the surroundings. Passive short video playback not only increases the user's operational burden but also lacks environmental adaptability.

[0003] Therefore, it is necessary to design a processing method and system for short video playback to solve the problems existing in the current technology. Summary of the Invention

[0004] In view of this, the present invention proposes a processing method and system for short video playback, aiming to solve the problems that current short video playback not only increases the user's operational burden, but also lacks adaptability to the environment.

[0005] In one aspect, the present invention proposes a processing method for short video playback, comprising:

[0006] Obtain user information and traverse the historical user database. Determine the reuse playback strategy or the model playback strategy based on the traversal results. When the model playback strategy is determined, obtain the sample interaction dataset and construct the strategy learning model based on the sample interaction dataset.

[0007] The user playback mode is determined based on the reuse playback strategy or the model playback strategy, playback environment data is obtained based on the user playback mode, and the playback environment data is preprocessed to determine the target playback environment data.

[0008] The target playback environment data is analyzed to generate playback adaptation tags. Based on the playback adaptation tags, all playback environment sub-data of the same type are analyzed and a playback environment influence sequence is constructed. According to the relationship between the standard playback environment influence sequence and the playback environment influence sequence, a playback hybrid influence set is constructed.

[0009] The association results of the playback mixed influence set are determined according to the association rule algorithm. The data in the playback mixed influence set that have not been associated and the association results are substituted into the Bayesian model to determine the playback improvement strategy. Based on the playback improvement strategy, the improvement resource energy consumption is determined, and the playback resource energy consumption of the short video is determined. Based on the relationship between the playback resource energy consumption and the improvement resource energy consumption, it is determined whether to adjust the playback improvement strategy, and the short video is played according to the adjustment result.

[0010] Furthermore, when determining the reuse playback strategy or model playback strategy based on the traversal results, the following are included:

[0011] The historical user database includes several historical user information entries and several historical user playback modes, and each historical user information entry corresponds to a historical user playback mode.

[0012] When the historical user database contains historical user information that matches the user information, it is determined to be the reuse playback strategy;

[0013] If no historical user information matching the user information exists in the historical user database, then the model playback strategy is determined to be correct.

[0014] Furthermore, when the model playback strategy is determined, acquiring the sample interaction dataset and constructing the strategy learning model based on the sample interaction dataset includes:

[0015] The sample interaction dataset is sampled proportionally to obtain the training set and the test set;

[0016] A Transformer model is pre-selected, and the Transformer model is iteratively trained based on the training set. The iteratively trained Transformer model is then validated based on the test set.

[0017] If the validation value of the Transformer model after the current iteration is less than the validation value of the Transformer model after the previous iteration, the learning rate of the Transformer model is adjusted and iterative training continues until the validation value of the Transformer model after the current iteration is greater than or equal to the validation value of the Transformer model after the previous iteration.

[0018] If the validation value of the Transformer model after the current iteration is greater than or equal to the validation value of the Transformer model after the previous iteration, then the iterative training is stopped, and the Transformer model after the iterative training is determined as the policy learning model.

[0019] Furthermore, when determining the user playback mode based on the reuse playback strategy or the model playback strategy, the following steps are included:

[0020] When the reuse playback strategy is determined, the historical user playback mode corresponding to the historical user information that is consistent with the user information in the historical user database is determined as the user playback mode.

[0021] When the model playback strategy is determined, the user information is substituted into the strategy learning model to determine the user playback mode.

[0022] Furthermore, when obtaining playback environment data based on the user playback mode and preprocessing the playback environment data to determine the target playback environment data, the process includes:

[0023] The playback environment data includes network environment data, environmental audio data, and several environmental angle images;

[0024] Image processing is performed on several environmental angle images, including denoising and geometric correction. Feature points are extracted from the processed environmental angle images based on the SIFT algorithm. The extracted feature points are matched based on the FLANN matching algorithm to determine the relevant information between the environmental angle images. The relevant information includes spatial position information and spatial rotation relationship. The environmental angle images are registered based on the relevant information. The registered environmental angle images are then merged according to the Poisson fusion algorithm to determine the target environmental image.

[0025] The network environment data is anomaly detected based on the three Sigma algorithm, and network fluctuation data is removed. The missing values ​​of the network environment data after removing network fluctuation data are filled based on the linear interpolation algorithm. The filled network environment data is then window-smoothed to determine the target network environment data.

[0026] The environmental audio data is denoised, the sampling rate of the denoised environmental audio data is unified, and the environmental audio data with unified sampling rate is framed, windowed and normalized to determine the time domain audio data. The time domain features of the time domain audio data are converted into frequency domain features based on short time Fourier transform to determine the target environmental audio data.

[0027] The target playback environment data includes the target environment image, the target network environment data, and the target environment audio data.

[0028] Furthermore, when analyzing the target playback environment data to generate playback adaptation tags, and analyzing all similar playback environment sub-data based on the playback adaptation tags to construct a playback environment influence series, the process includes:

[0029] Obtain qualified target playback environment data corresponding to the target playback environment data;

[0030] The playback adaptation markers include playback conformance adaptation markers and playback loss adaptation markers;

[0031] When the target environment image, target network environment data, and target environment audio data are all equal to the qualified target playback environment data, a playback compatibility adaptation flag is generated; otherwise, a playback mismatch adaptation flag is generated.

[0032] When the playback loss adaptation flag is generated, a playback environment influence sequence is constructed from all target playback environment sub-data of the same type in the target playback environment data.

[0033] Furthermore, when constructing the playback mixed influence set based on the standard playback environment influence sequence and the relationship between the playback environment influence sequence, it includes:

[0034] Obtain the standard playback environment influence sequence corresponding to the playback environment influence sequence, wherein the target playback environment sub-data in the playback environment influence sequence corresponds one-to-one with the standard target playback environment sub-data in the standard playback environment influence sequence;

[0035] Extract the target playback environment sub-data that is not equal to the standard playback environment influence sequence from the playback environment influence sequence and construct the playback mixed influence set.

[0036] Furthermore, when determining the association result of the playback mixed effect set according to the association rule algorithm, it includes:

[0037] The combined records of the target playback environment sub-data in the playback mixed influence set are transformed into a transaction dataset based on the association rule algorithm. Frequent itemsets are determined according to the support of the transaction dataset. Candidate association rules are generated in the frequent itemsets. The confidence of each candidate association rule is determined. Valid association rules that meet the minimum confidence are selected to determine the association result of the playback mixed influence set.

[0038] Furthermore, when determining whether to adjust the playback improvement strategy based on the relationship between the playback resource energy consumption and the improved resource energy consumption, the following steps are included:

[0039] If the energy consumption of the playback resources is greater than the energy consumption of the improved resources, it is determined that the playback improvement strategy will not be adjusted.

[0040] When the energy consumption of the playback resource is less than or equal to the energy consumption of the improved resource, it is determined that the playback improvement strategy should be adjusted, and the adjustment should be carried out in the order of short video image, short video sound and short video subtitle.

[0041] Compared with existing technologies, the beneficial effects of this invention are as follows: By traversing the historical user database to determine a reuse playback strategy or a model playback strategy, and constructing a strategy learning model when determining the model playback strategy, the invention accurately captures the operational preferences of different users, effectively solving the problem of misalignment between short video playback and user behavior habits. This reduces the burden on users to change their behavior habits and learn, enabling short videos to meet their own habitual playback needs, and improving the flexibility and dynamic adaptability of short video playback. In terms of environmental adaptation, the invention generates adaptation tags and constructs a playback environment influence sequence, and determines playback improvement strategies through association rule algorithms and Bayesian models, and adjusts them in conjunction with resource energy consumption relationships. This achieves optimized resource allocation, avoids unnecessary energy waste, and ensures high-quality short video playback while also considering energy-saving needs. This makes short video playback more flexible, improves the intelligence level of short video playback, and enhances the accuracy of dynamic adaptation to complex environments.

[0042] On the other hand, this application also provides a processing system for short video playback, for applying the above-described processing method for short video playback, including:

[0043] The data acquisition and processing module is configured to acquire user information and traverse the historical user database, determine the reuse playback strategy or the model playback strategy based on the traversal results, and when the model playback strategy is determined, acquire the sample interaction dataset and construct a strategy learning model based on the sample interaction dataset.

[0044] The first environmental processing module is configured to determine the user playback mode based on the reuse playback strategy or the model playback strategy, obtain playback environment data based on the user playback mode, and preprocess the playback environment data to determine the target playback environment data.

[0045] The second environmental processing module is configured to analyze the target playback environment data to generate playback adaptation tags, analyze all similar playback environment sub-data based on the playback adaptation tags and construct a playback environment influence sequence, and construct a playback hybrid influence set according to the relationship between the standard playback environment influence sequence and the playback environment influence sequence.

[0046] The playback processing module is configured to determine the association results of the playback mixed influence set according to the association rule algorithm, substitute the unassociated data in the playback mixed influence set and the association results into a Bayesian model to determine a playback improvement strategy, determine the improvement resource energy consumption based on the playback improvement strategy, determine the playback resource energy consumption of the short video playback, determine whether to adjust the playback improvement strategy based on the relationship between the playback resource energy consumption and the improvement resource energy consumption, and play the short video according to the adjustment result.

[0047] It is understandable that the above-mentioned processing method and system for short video playback have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0048] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0049] Figure 1 A flowchart illustrating a processing method for short video playback provided in an embodiment of the present invention;

[0050] Figure 2 This is a functional block diagram of a processing system for short video playback provided in an embodiment of the present invention. Detailed Implementation

[0051] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] See Figure 1 As shown in some embodiments of this application, a processing method for short video playback includes:

[0053] S100: Obtain user information and traverse the historical user database. Determine the reuse playback strategy or the model playback strategy based on the traversal results. When the model playback strategy is determined, obtain the sample interaction dataset and construct the strategy learning model based on the sample interaction dataset.

[0054] S200: Determine the user playback mode based on the reuse playback strategy or model playback strategy, obtain playback environment data based on the user playback mode, and preprocess the playback environment data to determine the target playback environment data.

[0055] S300: Analyze the target playback environment data to generate playback adaptation tags. Based on the playback adaptation tags, analyze all similar playback environment sub-data and construct a playback environment influence series. Based on the relationship between the standard playback environment influence series and the playback environment influence series, construct a playback mixed influence set.

[0056] S400: Determine the association results of the playback mixed influence set according to the association rule algorithm, substitute the data in the playback mixed influence set that have not been associated and the association results into the Bayesian model, determine the playback improvement strategy, determine the improvement resource energy consumption based on the playback improvement strategy, determine the playback resource energy consumption of short video playback, determine whether to adjust the playback improvement strategy based on the relationship between playback resource energy consumption and improvement resource energy consumption, and play the short video according to the adjustment result.

[0057] Specifically, user information is acquired and the historical user database is traversed to determine the playback strategy. User information includes identity information and basic behavioral tags. Identity information includes user registration account, mobile phone number, email-linked ID, device unique identifier, etc., used to accurately identify the user. Basic behavioral tags include operating habits (such as pause / play trigger methods: tapping the center of the screen, double-tapping the screen, pressing the volume button, etc.), swiping operation preferences (the speed of swiping up and down to switch videos, the extent of swiping left and right to adjust the progress), and the frequency of using full-screen mode (whether full-screen viewing is the default). These basic behavioral tags reflect the user's operating habits for video playback. By traversing the historical user database using user information, the model can quickly adapt to the video playback habits of existing users and enable the model's playback strategy for new users. This is achieved using a sample interaction dataset (containing up to...). Data on the click-pause method, swiping speed, and other behavioral habits of at least 100 users was collected to build a strategy learning model. This model was trained to address the diverse and volatile user habits, resolving the mismatch between fixed adjustment methods and user habits. This ensured that each user could adjust the short video playback according to their behavioral habits. After determining the user's playback mode (the short video playback operation mode determined by user information), miniature ambient light and sound sensors were used to acquire playback environment data, such as ambient light intensity and ambient sound decibels. Since device noise may affect the acquisition of playback environment data, it was preprocessed after acquisition to lay the foundation for accurate adaptation to different environments.

[0058] Specifically, playback adaptation tags are generated for the preprocessed target environment data to accurately assess whether the surrounding environment meets the requirements for short video playback. For example, in a strong light environment, is it suitable for direct playback of short videos? Based on this, a playback environment influence series is constructed, which covers all similar playback environment sub-data, such as the rate of change of ambient light (the sudden increase in light when moving from indoors to outdoors, the gradual decrease in light due to cloud cover on a cloudy day), and the spectral characteristics of light (full spectrum data of natural light, blue-biased spectrum data of LED lights), thus clearly presenting the dynamic influence between this type of data and short video playback. This system provides data support for subsequent short video playback and environmental adaptation. It constructs a playback mixed influence set based on the relationship between the standard playback environment influence series and the playback environment influence series. The standard playback environment influence series can be determined based on the short video player, playback experiment data, etc. For example, for the player screen's color accuracy parameters such as sRGB color gamut coverage, the optimal color temperature adjustment value under different ambient light spectra, such as natural light, fluorescent lamps, and incandescent lamps, is measured to form the standard playback environment influence series. When a short video is played in a reddish spectrum environment, the screen color temperature corresponding to the standard playback environment influence series should be 5000K to balance visual deviations. The relationship between environmental impact sequences and playback environmental impact sequences is used to construct a playback hybrid impact set. This set reflects the specific requirements of the actual scene for short video playback, accurately capturing the particularities of the surrounding environment and providing specific directions for targeted adjustments to subsequent playback improvement strategies. Association rule algorithms are used to determine the association results of the hybrid impact set. Unassociated data and associated results are then substituted into a Bayesian model to determine playback improvement strategies. Association rule algorithms can uncover the associations within the hybrid impact set. For example, when the rate of change in light is a sudden increase with fluctuations of approximately three levels within 10 seconds, the probability that the high-frequency band in the noise frequency distribution accounts for more than 60% is likely. The correlation between the two is strong, manifested by the rate of increase and the proportion of high-frequency noise. The Bayesian model can integrate these correlation results with uncorrelated data to generate playback improvement strategies, such as brightening or increasing the volume, to solve the passive adjustment of short video playback. This enables short video playback to actively adapt to the environment. Based on the playback improvement strategy, the energy consumption of improvement resources is determined. Improvement resource energy consumption includes the energy consumption of the backlight circuit and the computational energy consumption of the audio decoder. The relationship between playback resource energy consumption and improvement resource energy consumption determines whether to adjust the playback improvement strategy. This balances resource consumption while ensuring the adjustment of short video playback, thus ensuring efficient processing of short video playback.

[0059] Understandably, determining the user's playback mode based on a reused playback strategy or a model playback strategy solves the problem of mismatched user habits, enabling short video playback to automatically adapt to the operational preferences of different users. This allows for better control over short video playback, and the short video can automatically adjust in various scenarios to avoid issues such as glare, inaudibility, or interference with others. Whether it's a change in user habits or a sudden change in the environment, the short video playback can be adjusted accordingly in a timely manner, ensuring the flexibility and adaptability of short video playback.

[0060] In some embodiments of this application, when determining a reuse playback strategy or a model playback strategy based on the traversal results, the following steps are taken: the historical user database includes several historical user information and several historical user playback modes, and each historical user information corresponds to a historical user playback mode. When there is historical user information in the historical user database that matches the user information, it is determined to be a reuse playback strategy. When there is no historical user information in the historical user database that matches the user information, it is determined to be a model playback strategy.

[0061] Specifically, the historical user database stores several historical user information and historical user playback modes in a one-to-one correspondence. After obtaining the current user information, it is traversed and compared in the historical user database. If there is historical user information that is completely consistent with the current user information, it is determined to be a reuse playback strategy. The reuse playback strategy can directly call existing user behavior habits, avoiding repeated calculations for old users and reducing the computing power consumption and response latency during short video playback. If no matching historical user information is found after traversal, it may be a brand new user or the user information has changed. In this case, it is determined to be a model playback strategy. The model learns to dynamically match the user information's operating habits, so that short video playback can also obtain an adapted playback mode when facing new users, ensuring the flexibility and adaptability of short video playback.

[0062] In some embodiments of this application, when the model playback strategy is determined, the process of obtaining a sample interaction dataset and constructing a strategy learning model based on the sample interaction dataset includes: sampling the sample interaction dataset proportionally to obtain a training set and a test set; pre-selecting a Transformer model; iteratively training the Transformer model based on the training set; validating the iteratively trained Transformer model based on the test set; if the validation value of the Transformer model after the current iteration is less than the validation value of the Transformer model after the previous iteration, adjusting the learning rate of the Transformer model and continuing iterative training until the validation value of the Transformer model after the current iteration is greater than or equal to the validation value of the Transformer model after the previous iteration; if the validation value of the Transformer model after the current iteration is greater than or equal to the validation value of the Transformer model after the previous iteration, stopping the iterative training and determining the iteratively trained Transformer model as the strategy learning model.

[0063] Specifically, the sample interaction dataset contains behavioral data such as click-pause methods and swiping speeds from at least 100 users. This dataset is then sampled proportionally to create training and test sets, typically in a 7:3 ratio, to ensure both sets encompass various scenarios and guarantee the model's generalization ability. The training set is used to train the Transformer model, while the test set is used to evaluate its performance. The Transformer model comprises a feedforward neural network, multiple encoder layers, and multiple decoder layers, designed to capture complex relationships within the data. Using the Transformer model as the initial model, iterative training is performed using data from the training set. In each iteration, the model attempts to learn the relationships within the data. The Transformer model is trained to identify patterns and relationships to improve its predictive or classification capabilities. After each training iteration, the model is validated using data from the test set. Validation metrics include accuracy and recall to measure model performance. If the validation value of the model after the current iteration is lower than the previous validation value, it indicates a declining performance trend. In this case, the learning rate of the Transformer model needs to be adjusted, and iterative training continues to ensure the model stably approaches the global optimum. If the validation value of the model after the current iteration is greater than or equal to the previous validation value, it indicates improved or stable performance. At this point, iterative training stops, and the model is considered to have reached a satisfactory performance level. The iteratively trained Transformer model is then designated as the policy learning model. By training the Transformer model, the user playback patterns for each user can be accurately output, improving the flexibility and adaptability of short video playback.

[0064] In some embodiments of this application, when determining the user playback mode based on a reused playback strategy or a model playback strategy, the method includes: when the reused playback strategy is determined, the historical user playback mode corresponding to the historical user information that is consistent with the user information in the historical user database is determined as the user playback mode; when the model playback strategy is determined, the user information is substituted into the strategy learning model to determine the user playback mode.

[0065] Specifically, when using a reused playback strategy, historical user information that perfectly matches the current user information is extracted from the historical user database. The corresponding historical user playback patterns (including operating habits and preferences) are directly retrieved and used as the current user's playback pattern. When using a model-based playback strategy, since there is no matching historical data, the user information is input into the strategy learning model. The model analyzes the correlation between users and playback patterns to generate a personalized playback pattern that matches the user's characteristics—the user's playback pattern. The reused playback strategy directly reuses verified user habit data, avoiding repeated learning processes for existing users and reducing the computational burden during short video playback. The model-based playback strategy, on the other hand, dynamically constructs user playback patterns adapted to the behavioral habits of new users or users whose habits have not been recorded, ensuring that short video playback can adapt to each user's operating habits and guaranteeing the flexibility and adaptability of short video playback.

[0066] In some embodiments of this application, when acquiring playback environment data based on the user's playback mode and preprocessing the playback environment data to determine the target playback environment data, the process includes: the playback environment data includes network environment data, ambient audio data, and several ambient angle images; image processing is performed on the several ambient angle images, including denoising and geometric correction; feature points are extracted from the processed ambient angle images using the SIFT algorithm; the extracted feature points are matched using the FLANN matching algorithm; relevant information between the ambient angle images is determined, including spatial position information and spatial rotation relationships; the ambient angle images are registered based on the relevant information; and the registered ambient angle images are then processed using the Poisson fusion algorithm. The target environment image is determined by merging. Outlier detection is performed on the network environment data based on the three Sigma algorithm, and network fluctuation data is removed. Missing values ​​in the network environment data after removing network fluctuation data are filled using a linear interpolation algorithm. The filled network environment data is then windowed and smoothed to determine the target network environment data. The environmental audio data is denoised. The denoised environmental audio data is then standardized at a uniform sampling rate. The standardized environmental audio data is then framed, windowed, and normalized to determine the temporal audio data. The temporal features of the temporal audio data are converted into frequency features based on the short-time Fourier transform to determine the target environmental audio data. The target playback environment data includes the target environment image, the target network environment data, and the target environmental audio data.

[0067] Specifically, differentiated processing is applied to the three types of data in the playback environment data. For several environmental angle images, each environmental angle image represents an image taken from one angle of the environment, and several environmental angle images together constitute the ambient lighting information during short video playback. Denoising and geometric correction are used to eliminate shooting noise and lens distortion, ensuring image quality. Then, the SIFT algorithm is used to extract scale-invariant feature points, and the FLANN matching algorithm is used to locate common features of images from different angles, thereby determining the spatial position and rotation relationship between images, achieving accurate matching of images from multiple perspectives. The Poisson fusion algorithm is used to merge the registered images to generate a target environment image that fully reflects the panoramic view of the environment, solving the problems of limited perspective of single images and distortion of multi-image stitching, thus fully restoring the visual features of the environment. For network environment data, the three sigma algorithm is used to identify and remove outliers that exceed the normal fluctuation range (such as instantaneous network outages and sudden high latency), and missing values ​​in the data acquisition gaps are filled by linear interpolation. Then, window smoothing is used to filter high-frequency fluctuations, thereby eliminating unstable data in the network data, and thus obtaining smooth and stable target network environment data, which accurately reflects the actual network quality during short video playback. For environmental audio data, noise reduction is first performed to remove background noise. Environmental audio data collected by different devices may have different default sampling rates. If these are directly mixed, it will lead to differences in data volume and time axis scale for audio of the same duration, which in turn will cause problems such as chaotic time granularity during frame windowing and inconsistent frequency resolution during frequency domain conversion. By unifying the sampling rate, the noise-reduced environmental audio data can be mapped to the same time-frequency coordinate system, ensuring that operations such as short-time Fourier transform are performed on a uniform scale, avoiding feature distortion caused by format differences. Then, frame windowing and normalization processing are used to convert continuous audio into analyzable time domain segments. Finally, short-time Fourier transform is used to convert time domain features into frequency domain features to generate target environmental audio data, thus realizing noise reduction and feature extraction of environmental audio data.

[0068] It is understandable that playback environment data may contain issues such as noise, limited viewing angles, and format differences. Direct use of such data may lead to misjudgments. However, targeted processing can unify data standards, eliminate interference, and extract effective features, providing a reliable basis for environment adaptation. This allows the target environment data to accurately reflect the scene characteristics during actual playback, improving the accuracy of subsequent adaptation.

[0069] In some embodiments of this application, when analyzing target playback environment data to generate playback adaptation tags, and analyzing all similar playback environment sub-data based on the playback adaptation tags to construct a playback environment influence sequence, the process includes: obtaining qualified target playback environment data corresponding to the target playback environment data; the playback adaptation tags include playback conformance adaptation tags and playback loss adaptation tags; when the target environment image, target network environment data, and target environment audio data are all equal to qualified target playback environment data, a playback conformance adaptation tag is generated; otherwise, a playback loss adaptation tag is generated; when a playback loss adaptation tag is generated, a playback environment influence sequence is constructed from all similar target playback environment sub-data in the target playback environment data.

[0070] Specifically, the qualified target playback environment data is determined through experimental data and the usage scenario requirements of the player. Qualified target playback environment data includes qualified target environment images, qualified target network environment data, and qualified target environment audio data. By comparing against a benchmark, the focus is on the scene features that need optimization. Qualified target playback environment data covers standards such as image clarity, network stability, and audio signal-to-noise ratio. When all target playback environment data meet the standards, a playback compatibility flag is generated, indicating that the short video playback requires no special adjustment for the current environment and can be played directly. If any type of data in the target playback environment deviates from the benchmark, a playback misfit flag is generated, indicating that the short video playback is not suitable for the current environment. If targeted adjustments are needed, all similar target playback environment sub-data are constructed into a playback environment influence series. For example, for target environment images, the target playback environment sub-data includes the rate of change of ambient light and the spectral characteristics of light, thus clearly presenting the dynamic influence relationship between this type of data and short video playback. Using qualified target playback environment data as a benchmark, normal and potentially abnormal environments can be quickly distinguished, avoiding repeated analysis of well-adapted scenes. The loss of adaptation markers indicates that optimization and adjustment are needed for specific environments. Constructing a playback environment influence series through similar target playback environment sub-data can focus on the common patterns of similar problems, providing a precise basis for subsequent adjustments and ensuring the adaptability of short video playback to the environment.

[0071] In some embodiments of this application, when constructing a playback mixed influence set based on the relationship between the standard playback environment influence sequence and the playback environment influence sequence, the process includes: obtaining the standard playback environment influence sequence corresponding to the playback environment influence sequence, wherein the target playback environment sub-data in the playback environment influence sequence corresponds one-to-one with the standard target playback environment sub-data in the standard playback environment influence sequence, extracting the target playback environment sub-data in the playback environment influence sequence that is not equal to the standard playback environment influence sequence, and constructing the playback mixed influence set.

[0072] Specifically, the standard playback environment influence sequence is a benchmark sequence based on universal adaptation rules, containing ideal parameters for various target playback environment sub-data. The playback environment influence sequence is a feature sequence of the current actual environment. The target playback environment sub-data in the playback environment influence sequence corresponds one-to-one with the standard target playback environment sub-data in the standard playback environment influence sequence, so that each target playback environment sub-data in the playback environment influence sequence can find a corresponding benchmark for comparison, thereby identifying the adaptation requirements of short video playback and ensuring the adaptability of short video playback to the environment. By comparing one by one, the target playback environment sub-data that does not match the standard playback environment influence sequence (i.e., environmental features that deviate from the ideal state) is extracted from the playback environment influence sequence, and these discrepancies are integrated into a playback hybrid influence set. This focuses on the environmental variables that need to be adjusted for short videos, improving the targeting and accuracy of environment-adaptive playback, so that subsequent playback improvement strategies can fit the actual adjustment needs, ensuring that the playback of short videos always maintains optimal matching with the current environment.

[0073] In some embodiments of this application, when determining the association result of the playback mixed effect set according to the association rule algorithm, the method includes: converting the combined records of the target playback environment sub-data in the playback mixed effect set into a transaction dataset based on the association rule algorithm, determining the frequent itemset according to the support of the transaction dataset, generating candidate association rules in the frequent itemset, determining the confidence of each candidate association rule, and selecting the valid association rule that meets the minimum confidence to determine the association result of the playback mixed effect set.

[0074] Specifically, the combined records of the target playback environment sub-data in the playback mixed influence set are transformed into a transaction dataset, making the scattered environmental difference information conform to the algorithm's processing format. Then, frequent data combinations (frequent itemsets) are selected through support filtering, focusing on candidate objects with high probability of association. Subsequently, candidate association rules are generated based on frequent itemsets, and the reliability of the rules is quantified by confidence. Finally, valid rules that meet the minimum confidence level are retained as association results. Support and minimum support can be dynamically adjusted according to the amount of data in the target playback environment sub-data in the playback mixed influence set. The data in the playback mixed influence set have certain special characteristics, and there may be implicit associations (such as the co-occurrence pattern of specific light and noise). The association rule algorithm, through dual filtering of support and confidence, can both eliminate random meaningless combinations and accurately capture statistically frequent association patterns, avoiding bias caused by subjective judgment. It transforms the originally scattered target playback environment sub-data into directly applicable association rules, providing data support for playback improvement strategies. By mining the inherent connections, the playback of short videos becomes more flexible, improving the intelligence of short video playback and the accuracy of dynamic adaptation to complex environments.

[0075] In some embodiments of this application, when determining whether to adjust the playback improvement strategy based on the relationship between playback resource energy consumption and improvement resource energy consumption, the following steps are taken: when the playback resource energy consumption is greater than the improvement resource energy consumption, it is determined that the playback improvement strategy will not be adjusted; when the playback resource energy consumption is less than or equal to the improvement resource energy consumption, it is determined that the playback improvement strategy will be adjusted, and the adjustment is performed in the order of short video image, short video sound and short video subtitle.

[0076] Specifically, the association rule algorithm mines the correlations between sub-data of the target playback environment from the playback mixed influence set, while retaining uncorrelated data (i.e., isolated or weakly correlated environmental features). Then, the correlation results (known patterns) and uncorrelated data (uncertainty information) are jointly input into a Bayesian model. Utilizing the Bayesian model's modeling of uncertainty, it integrates prior knowledge and real-time data to generate probabilistic playback improvement strategies, such as predicting optimal brightness and volume parameters based on environmental features. The training process of the Bayesian model is consistent with that of the Transformer model, and will not be repeated here. By quantifying the resource consumption of the playback improvement strategy (improvement resource energy consumption), including the energy consumption of the backlight circuit and the computational energy consumption of the audio decoder, as well as the actual consumption required when playing the short video (playback resource energy consumption), if the playback resource energy consumption is greater than the improved resource energy consumption, it indicates that the short video playback can meet the adjustment requirements for the environment. In this case, no adjustment is made to the playback improvement strategy, and the playback of the short video (such as brightness and volume) is adjusted according to the playback improvement strategy. When the playback resource energy consumption is less than or equal to the improved resource energy consumption, it indicates that the short video playback does not meet the adjustment requirements for the environment. Here, "not meeting" does not mean that the playback function is malfunctioning, but rather that the current playback improvement strategy cannot respond optimally to environmental changes to achieve the best performance for the short video. Playback resource energy consumption represents the energy consumption of short video playback, while improved resource energy consumption is the pre-optimized energy consumption after algorithm optimization. When the investment is substantial and resource consumption for improvement is high, the playback improvement strategy will be adjusted in the order of short video image, short video sound, and short video subtitles. For example, if the playback improvement strategy for dark environments is to increase brightness by two levels, then when adjusting the playback improvement strategy, one level will be reduced accordingly to balance the relationship between playback resource energy consumption and improvement resource energy consumption. The parameters in the playback improvement strategy will be reduced sequentially according to the order of picture quality, sound volume, and subtitle brightness to avoid excessive resource consumption in pursuit of experience optimization. This forms a closed loop of data mining, strategy generation, and effect verification, enabling the playback improvement strategy to dynamically iterate with the environment and resource status, thereby adapting to changes in short video playback and actual scenarios, improving the intelligence of short video playback and the accuracy of dynamic adaptation to complex environments.

[0077] In summary, the beneficial effects of this invention are as follows: By traversing the historical user database to determine a reuse playback strategy or a model playback strategy, and constructing a strategy learning model when determining the model playback strategy, the invention accurately captures the operational preferences of different users, effectively solving the problem of misalignment between short video playback and user behavior habits. This reduces the burden on users to change their behavior habits and learn, enabling short videos to meet their own habitual playback needs, thus improving the flexibility and dynamic adaptability of short video playback. In terms of environmental adaptation, the invention generates adaptation tags and constructs a playback environment influence sequence, and determines playback improvement strategies through association rule algorithms and Bayesian models, adjusting them in conjunction with resource energy consumption relationships. This achieves optimized resource allocation, avoiding unnecessary energy waste, ensuring high-quality short video playback while also considering energy-saving needs, making short video playback flexible, improving the intelligence level of short video playback, and enhancing the dynamic adaptation accuracy to complex environments.

[0078] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this embodiment provides a processing system for short video playback, which applies the above-described processing method for short video playback, including:

[0079] The data acquisition and processing module is configured to acquire user information and traverse the historical user database. Based on the traversal results, it determines the reuse playback strategy or the model playback strategy. When the model playback strategy is determined, it acquires the sample interaction dataset and constructs a strategy learning model based on the sample interaction dataset.

[0080] The first environmental processing module is configured to determine the user's playback mode based on a reused playback strategy or a model playback strategy, acquire playback environment data based on the user's playback mode, and preprocess the playback environment data to determine the target playback environment data.

[0081] The second environmental processing module is configured to analyze the target playback environment data to generate playback adaptation tags, analyze all similar playback environment sub-data based on the playback adaptation tags and construct a playback environment influence series, and construct a playback mixed influence set according to the relationship between the standard playback environment influence series and the playback environment influence series.

[0082] The playback processing module is configured to determine the association results of the playback mixed influence set according to the association rule algorithm, substitute the unassociated data and association results in the playback mixed influence set into the Bayesian model to determine the playback improvement strategy, determine the improvement resource energy consumption based on the playback improvement strategy, determine the playback resource energy consumption of short video playback, determine whether to adjust the playback improvement strategy based on the relationship between playback resource energy consumption and improvement resource energy consumption, and play the short video according to the adjustment result.

[0083] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0085] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A processing method for short video playback, characterized in that, include: Obtain user information and traverse the historical user database. Determine the reuse playback strategy or the model playback strategy based on the traversal results. When the model playback strategy is determined, obtain the sample interaction dataset and construct a strategy learning model based on the sample interaction dataset. The user playback mode is determined based on the reuse playback strategy or the model playback strategy, playback environment data is obtained based on the user playback mode, and the playback environment data is preprocessed to determine the target playback environment data. The target playback environment data is analyzed to generate playback adaptation tags. Based on the playback adaptation tags, all playback environment sub-data of the same type are analyzed and a playback environment influence sequence is constructed. According to the relationship between the standard playback environment influence sequence and the playback environment influence sequence, a playback hybrid influence set is constructed. The association results of the playback mixed influence set are determined according to the association rule algorithm. The data in the playback mixed influence set that have not been associated and the association results are substituted into the Bayesian model to determine the playback improvement strategy. Based on the playback improvement strategy, the improvement resource energy consumption is determined, and the playback resource energy consumption of the short video is determined. Based on the relationship between the playback resource energy consumption and the improvement resource energy consumption, it is determined whether to adjust the playback improvement strategy, and the short video is played according to the adjustment result. When determining the reuse playback strategy or model playback strategy based on the traversal results, the following are included: The historical user database includes several historical user information entries and several historical user playback modes, and each historical user information entry corresponds to a historical user playback mode. When the historical user database contains historical user information that matches the user information, it is determined to be the reuse playback strategy; If no historical user information matching the user information exists in the historical user database, then the model playback strategy is determined to be correct. When obtaining playback environment data based on the user playback mode and preprocessing the playback environment data to determine the target playback environment data, the process includes: The playback environment data includes network environment data, environmental audio data, and several environmental angle images; Image processing is performed on several environmental angle images, including denoising and geometric correction. Feature points are extracted from the processed environmental angle images based on the SIFT algorithm. The extracted feature points are matched based on the FLANN matching algorithm to determine the relevant information between the environmental angle images. The relevant information includes spatial position information and spatial rotation relationship. The environmental angle images are registered based on the relevant information. The registered environmental angle images are then merged according to the Poisson fusion algorithm to determine the target environmental image. The network environment data is anomaly detected based on the three Sigma algorithm, and network fluctuation data is removed. The missing values ​​of the network environment data after removing network fluctuation data are filled based on the linear interpolation algorithm. The filled network environment data is then window-smoothed to determine the target network environment data. The environmental audio data is denoised, the sampling rate of the denoised environmental audio data is unified, and the environmental audio data with unified sampling rate is framed, windowed and normalized to determine the time domain audio data. The time domain features of the time domain audio data are converted into frequency domain features based on short time Fourier transform to determine the target environmental audio data. The target playback environment data includes the target environment image, target network environment data, and target environment audio data; When analyzing the target playback environment data to generate playback adaptation tags, and analyzing all similar playback environment sub-data based on the playback adaptation tags to construct a playback environment influence series, the process includes: Obtain qualified target playback environment data corresponding to the target playback environment data; The playback adaptation markers include playback conformance adaptation markers and playback loss adaptation markers; When the target environment image, target network environment data, and target environment audio data are all equal to the qualified target playback environment data, a playback compatibility adaptation flag is generated; otherwise, a playback mismatch adaptation flag is generated. When the playback loss adaptation flag is generated, a playback environment influence sequence is constructed from all target playback environment sub-data of the same type in the target playback environment data.

2. The processing method for short video playback according to claim 1, characterized in that, When the model playback strategy is determined, the process of acquiring a sample interaction dataset and constructing a strategy learning model based on the sample interaction dataset includes: The sample interaction dataset is sampled proportionally to obtain the training set and the test set; A Transformer model is pre-selected, and the Transformer model is iteratively trained based on the training set. The iteratively trained Transformer model is then validated based on the test set. If the validation value of the Transformer model after the current iteration is less than the validation value of the Transformer model after the previous iteration, the learning rate of the Transformer model is adjusted and iterative training continues until the validation value of the Transformer model after the current iteration is greater than or equal to the validation value of the Transformer model after the previous iteration. If the validation value of the Transformer model after the current iteration is greater than or equal to the validation value of the Transformer model after the previous iteration, then the iterative training is stopped, and the Transformer model after the iterative training is determined as the policy learning model.

3. The processing method for short video playback according to claim 2, characterized in that, When determining the user playback mode based on the reuse playback strategy or the model playback strategy, the following is included: When the reuse playback strategy is determined, the historical user playback mode corresponding to the historical user information that is consistent with the user information in the historical user database is determined as the user playback mode. When the model playback strategy is determined, the user information is substituted into the strategy learning model to determine the user playback mode.

4. The processing method for short video playback according to claim 3, characterized in that, When constructing a playback hybrid influence set based on the standard playback environment influence sequence and the relationship between the playback environment influence sequence, the following are included: Obtain the standard playback environment influence sequence corresponding to the playback environment influence sequence, wherein the target playback environment sub-data in the playback environment influence sequence corresponds one-to-one with the standard target playback environment sub-data in the standard playback environment influence sequence; Extract the target playback environment sub-data that is not equal to the standard playback environment influence sequence from the playback environment influence sequence and construct the playback mixed influence set.

5. The processing method for short video playback according to claim 4, characterized in that, When determining the association result of the playback mixed effect set according to the association rule algorithm, it includes: The combined records of the target playback environment sub-data in the playback mixed influence set are transformed into a transaction dataset based on the association rule algorithm. Frequent itemsets are determined according to the support of the transaction dataset. Candidate association rules are generated in the frequent itemsets. The confidence of each candidate association rule is determined. Valid association rules that meet the minimum confidence are selected to determine the association result of the playback mixed influence set.

6. The processing method for short video playback according to claim 5, characterized in that, When determining whether to adjust the playback improvement strategy based on the relationship between the playback resource energy consumption and the improved resource energy consumption, the following steps are included: If the energy consumption of the playback resources is greater than the energy consumption of the improved resources, it is determined that the playback improvement strategy will not be adjusted. When the energy consumption of the playback resource is less than or equal to the energy consumption of the improved resource, it is determined that the playback improvement strategy should be adjusted, and the adjustment should be carried out in the order of short video image, short video sound and short video subtitle.

7. A processing system for short video playback, used to apply the processing method for short video playback as described in any one of claims 1-6, characterized in that, include: The data acquisition and processing module is configured to acquire user information and traverse the historical user database, determine the reuse playback strategy or the model playback strategy based on the traversal results, and when the model playback strategy is determined, acquire the sample interaction dataset and construct a strategy learning model based on the sample interaction dataset. The first environmental processing module is configured to determine the user playback mode based on the reuse playback strategy or the model playback strategy, obtain playback environment data based on the user playback mode, and preprocess the playback environment data to determine the target playback environment data. The second environmental processing module is configured to analyze the target playback environment data to generate playback adaptation tags, analyze all similar playback environment sub-data based on the playback adaptation tags and construct a playback environment influence sequence, and construct a playback hybrid influence set according to the relationship between the standard playback environment influence sequence and the playback environment influence sequence. The playback processing module is configured to determine the association results of the playback mixed influence set according to the association rule algorithm, substitute the data in the playback mixed influence set that has not been associated and the association results into a Bayesian model to determine a playback improvement strategy, determine the improvement resource energy consumption based on the playback improvement strategy, determine the playback resource energy consumption of short video playback, determine whether to adjust the playback improvement strategy based on the relationship between the playback resource energy consumption and the improvement resource energy consumption, and play the short video according to the adjustment result. When determining the reuse playback strategy or model playback strategy based on the traversal results, the following are included: The historical user database includes several historical user information entries and several historical user playback modes, and each historical user information entry corresponds to a historical user playback mode. When the historical user database contains historical user information that matches the user information, it is determined to be the reuse playback strategy; If no historical user information matching the user information exists in the historical user database, then the model playback strategy is determined to be correct. When obtaining playback environment data based on the user playback mode and preprocessing the playback environment data to determine the target playback environment data, the process includes: The playback environment data includes network environment data, environmental audio data, and several environmental angle images; Image processing is performed on several environmental angle images, including denoising and geometric correction. Feature points are extracted from the processed environmental angle images based on the SIFT algorithm. The extracted feature points are matched based on the FLANN matching algorithm to determine the relevant information between the environmental angle images. The relevant information includes spatial position information and spatial rotation relationship. The environmental angle images are registered based on the relevant information. The registered environmental angle images are then merged according to the Poisson fusion algorithm to determine the target environmental image. The network environment data is anomaly detected based on the three Sigma algorithm, and network fluctuation data is removed. The missing values ​​of the network environment data after removing network fluctuation data are filled based on the linear interpolation algorithm. The filled network environment data is then window-smoothed to determine the target network environment data. The environmental audio data is denoised, the sampling rate of the denoised environmental audio data is unified, and the environmental audio data with unified sampling rate is framed, windowed and normalized to determine the time domain audio data. The time domain features of the time domain audio data are converted into frequency domain features based on short time Fourier transform to determine the target environmental audio data. The target playback environment data includes the target environment image, target network environment data, and target environment audio data; When analyzing the target playback environment data to generate playback adaptation tags, and analyzing all similar playback environment sub-data based on the playback adaptation tags to construct a playback environment influence series, the process includes: Obtain qualified target playback environment data corresponding to the target playback environment data; The playback adaptation markers include playback conformance adaptation markers and playback loss adaptation markers; When the target environment image, target network environment data, and target environment audio data are all equal to the qualified target playback environment data, a playback compatibility adaptation flag is generated; otherwise, a playback mismatch adaptation flag is generated. When the playback loss adaptation flag is generated, a playback environment influence sequence is constructed from all target playback environment sub-data of the same type in the target playback environment data.

Images