ABR parameter adjusting method and device, electronic equipment, storage medium and program product

Abstract

The invention provides an ABR parameter adjusting method and device, electronic equipment, a storage medium and a program product. The adjusting method comprises the steps that current video playing feature information of a user is acquired, and the video playing feature information comprises network condition information, player state information and user playing behavior information; based on the video playing feature information, determining an optimal parameter value of a specific ABR parameter of the player, the specific ABR parameter including a parameter used for representing the tolerance of a user to jamming in an ABR algorithm; and updating the parameter value of the specific ABR parameter to the optimal parameter value so as to select the bit rate of the video stream by using the ABR algorithm with the updated parameter value.

Description

Technical Field

[0001] The present disclosure generally relates to the technical field of video stream bit rate, and more specifically, to an ABR parameter adjustment method, apparatus, electronic device, storage medium, and program product. Background Art

[0002] In recent years, with advances in network technology and the growing popularity of user-generated content, video has become an indispensable part of people's daily lives. Video applications are a major contributor to downstream traffic. Streaming systems are fundamentally designed to improve user Quality of Experience (QoE). In production environments, user engagement metrics, such as viewing time and dropout rates, are key indicators of QoE.

[0003] Adaptive bitrate (ABR) algorithms play a crucial role in user-level quality of experience (QoE) in online video streaming. These algorithms dynamically optimize the bitrate selection for upcoming video segments based on the current buffer state and historical bandwidth measurements, typically striving to minimize lag time while improving video quality and playback smoothness. While QoE represents the ultimate optimization goal, directly optimizing QoE metrics presents significant challenges. Therefore, QoE is often approximated through measurable Quality of Service (QoS) metrics, serving as proxy optimization targets.

[0004] However, the optimization method based on system-level QoS indicators has reached its performance limit. lin ) has limited improvement potential in production systems and may even have a negative impact on core QoE indicators such as viewing time. In addition, users have significant differences in QoS sensitivity, and a unified QoE lin The formula or the neural network fitting model based on Mean Opinion Scores (MOS) can only effectively represent the average user experience and cannot capture the individual specific QoE characteristics. Summary of the Invention

[0005] The exemplary embodiments of the present disclosure provide a method, apparatus, electronic device, storage medium, and program product for adjusting ABR parameters, which can solve at least one of the above-mentioned problems in the prior art.

[0006] According to a first aspect of an embodiment of the present disclosure, a method for adjusting ABR parameters is provided, comprising: obtaining current video playback feature information of a user, wherein the video playback feature information includes: network status information, player status information, and user playback behavior information; determining an optimal parameter value of a specific ABR parameter of the player based on the video playback feature information, wherein the specific ABR parameter includes: a parameter in an ABR algorithm used to characterize the user's tolerance for stuttering; and updating the parameter value of the specific ABR parameter to the optimal parameter value, so as to select the bit rate of the video stream using the ABR algorithm after the parameter value is updated.

[0007] Optionally, the specific ABR parameters further include: parameters in the ABR algorithm used to characterize a user's preference for smoothness, and / or parameters in the ABR algorithm used to characterize a user's preference for video quality.

[0008] Optionally, the network status information includes: bandwidth distribution; the player status information includes: bit rate, buffer status information; the user playback behavior information includes: freeze response information.

[0009] Optionally, determining the optimal parameter value of the specific ABR parameter of the player based on the video playback characteristic information includes: determining the optimal parameter value of the specific ABR parameter through an optimization algorithm based on the video playback characteristic information, wherein a goal of the optimization algorithm is to find a parameter value that can minimize the user's video exit probability.

[0010] Optionally, in the step of determining the optimal parameter value of the specific ABR parameter through an optimization algorithm based on the video playback feature information, the video exit probability corresponding to the candidate parameter value is determined in the following manner: based on the video playback feature information, simulating a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value; and taking the sum of the video exit probabilities corresponding to each of the series of video playback trajectories as the video exit probability corresponding to the candidate parameter value.

[0011] Optionally, based on the video playback feature information, simulating a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value, including: determining a current bandwidth value based on bandwidth distribution; obtaining a current bit rate selected by the ABR algorithm when the specific ABR parameter is the candidate parameter value based on the current bandwidth value and current buffer state information; simulating a virtual playback trajectory of the jth video segment of the ith video according to the current bit rate, wherein the initial values ​​of i and j are 1; simulating the virtual playback trajectory corresponding to the jth video segment of the ith video based on the current bit rate. The freeze time, video quality, smoothness, and historical freeze reaction information are used to predict the exit probability of the j+1th video segment of the i-th video, and based on the exit probability, determine whether to continue playing the j+1th video segment of the i-th video; if it is determined to continue playing the j+1th video segment of the i-th video, set j=j+1 and return to execute the method of determining the current bandwidth value based on the bandwidth distribution; if it is determined not to continue playing the j+1th video segment of the i-th video, use the exit probability as the video exit probability corresponding to the i-th video playback track, set i=i+1 and j=1 and return to execute the method of determining the current bandwidth value based on the bandwidth distribution.

[0012] Optionally, the predicting the exit probability of the j+1th video segment of the ith video includes: predicting a first exit probability of the j+1th video segment of the ith video based on the current bit rate, the current bandwidth value, the freeze time corresponding to the virtual playback trajectory of the jth video segment of the ith video, and historical freeze reaction information; predicting a second exit probability of the j+1th video segment of the ith video based on the smoothness corresponding to the virtual playback trajectory of the jth video segment of the ith video; predicting a third exit probability of the j+1th video segment of the ith video based on the video quality corresponding to the virtual playback trajectory of the jth video segment of the ith video; and determining the exit probability of the j+1th video segment of the ith video based on the first exit probability, the second exit probability, and the third exit probability.

[0013] Optionally, predicting the first exit probability of the j+1th video segment of the i-th video includes: inputting time series data into a pre-trained neural network model to obtain the first exit probability output by the neural network model; wherein, the time series data includes time series data of the following dimensions: bit rate, bandwidth value, freeze time, interval from the last freeze, and interval from the last freeze that caused user exit.

[0014] Optionally, obtaining the user's current video playback feature information includes: in response to the occurrence of a freeze condition meeting a preset condition, obtaining the user's current video playback feature information.

[0015] According to a second aspect of an embodiment of the present disclosure, an ABR parameter adjustment device is provided, including: a feature information acquisition unit, configured to obtain a user's current video playback feature information, wherein the video playback feature information includes: network status information, player status information, and user playback behavior information; an optimization unit, configured to determine an optimal parameter value of a specific ABR parameter of the player based on the video playback feature information, wherein the specific ABR parameter includes: a parameter in the ABR algorithm used to characterize the user's tolerance for lag; a parameter value updating unit, configured to update the parameter value of the specific ABR parameter to the optimal parameter value, so as to use the ABR algorithm after the parameter value is updated to select the bit rate of the video stream.

[0016] Optionally, the specific ABR parameters further include: parameters in the ABR algorithm used to characterize a user's preference for smoothness, and / or parameters in the ABR algorithm used to characterize a user's preference for video quality.

[0017] Optionally, the network status information includes: bandwidth distribution; the player status information includes: bit rate, buffer status information; the user playback behavior information includes: freeze response information.

[0018] Optionally, the optimization unit is configured to: determine the optimal parameter value of the specific ABR parameter through an optimization algorithm based on the video playback feature information, wherein the goal of the optimization algorithm is to find a parameter value that can minimize the user's video exit probability.

[0019] Optionally, the optimization unit is configured to determine the video exit probability corresponding to the candidate parameter value in the following manner: based on the video playback feature information, simulate a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value; and take the sum of the video exit probabilities corresponding to each of the series of video playback trajectories as the video exit probability corresponding to the candidate parameter value.

[0020] Optionally, the optimization unit is configured to: determine a current bandwidth value based on a bandwidth distribution; obtain a current bit rate selected by the ABR algorithm when the specific ABR parameter is the candidate parameter value based on the current bandwidth value and current buffer status information; simulate a virtual playback trajectory of the jth video segment of the ith video according to the current bit rate, where the initial values ​​of i and j are 1; predict an exit probability of the j+1th video segment of the ith video based on the freeze time, video quality, smoothness, and historical freeze response information corresponding to the virtual playback trajectory of the jth video segment of the ith video, and determine whether to continue playing the j+1th video segment of the ith video based on the exit probability; if it is determined to continue playing the j+1th video segment of the ith video, set j=j+1 and return to executing the determination of the current bandwidth value based on the bandwidth distribution; if it is determined not to continue playing the j+1th video segment of the ith video, use the exit probability as the video exit probability corresponding to the ith video playback trajectory, set i=i+1 and j=1 and return to executing the determination of the current bandwidth value based on the bandwidth distribution.

[0021] Optionally, the optimization unit is configured to: predict a first exit probability of the j+1th video segment of the ith video based on the current bit rate, the current bandwidth value, the freeze time corresponding to the virtual playback trajectory of the jth video segment of the ith video, and the historical freeze reaction information; predict a second exit probability of the j+1th video segment of the ith video based on the smoothness corresponding to the virtual playback trajectory of the jth video segment of the ith video; predict a third exit probability of the j+1th video segment of the ith video based on the video quality corresponding to the virtual playback trajectory of the jth video segment of the ith video; and determine the exit probability of the j+1th video segment of the ith video based on the first exit probability, the second exit probability, and the third exit probability.

[0022] Optionally, the optimization unit is configured to: input the time series data into a pre-trained neural network model to obtain a first exit probability output by the neural network model; wherein the time series data includes time series data of the following dimensions: bit rate, bandwidth value, freeze time, interval from the last freeze, and interval from the last freeze that caused user exit.

[0023] Optionally, the feature information acquisition unit is configured to: in response to the occurrence of a freeze condition meeting a preset condition, acquire the user's current video playback feature information.

[0024] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: at least one processor; and at least one memory storing computer-executable instructions, wherein the computer-executable instructions, when executed by the at least one processor, prompt the at least one processor to execute the ABR parameter adjustment method as described above.

[0025] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to execute the ABR parameter adjustment method as described above.

[0026] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising computer instructions, which, when executed by at least one processor, implement the ABR parameter adjustment method as described above.

[0027] According to the ABR parameter adjustment method, device, electronic device, storage medium and program product of the exemplary embodiments of the present disclosure, taking into account the significant individual differences and certain temporal dynamics of users' tolerance to lag, the user QoE is perceived by analyzing the user's natural playback behavior during video playback (such as the reaction to lag), and the key parameters of the ABR algorithm are dynamically adjusted accordingly to achieve the optimization goal of dynamically adjusting the ABR algorithm.

[0028] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute an improper limitation of the present disclosure.

[0030] Figure 1 A flowchart illustrating a method for adjusting ABR parameters according to an exemplary embodiment of the present disclosure is provided;

[0031] Figure 2 A flowchart illustrating a method for simulating a series of possible video playback trajectories starting from a current state of a player according to an exemplary embodiment of the present disclosure is shown;

[0032] Figure 3 A flowchart illustrating a method for predicting an exit probability of a j+1th video segment of an i-th video according to an exemplary embodiment of the present disclosure is shown;

[0033] Figure 4 A schematic diagram illustrating an ABR parameter adjustment system according to an exemplary embodiment of the present disclosure;

[0034] Figure 5 A schematic diagram illustrating a neural network model according to an exemplary embodiment of the present disclosure;

[0035] Figure 6 A structural block diagram of an ABR parameter adjustment device according to an exemplary embodiment of the present disclosure is shown;

[0036] Figure 7 A structural block diagram of an electronic device according to an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0037] In order to enable ordinary people in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0038] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure as detailed in the appended claims.

[0039] It should be noted that the phrase "at least one of the items" in this disclosure includes three types of parallel situations: "any one of the items", "a combination of any multiple items of the items", and "all of the items". For example, "including at least one of A and B" includes the following three parallel situations: (1) including A; (2) including B; (3) including A and B. For another example, "performing at least one of step 1 and step 2" includes the following three parallel situations: (1) performing step 1; (2) performing step 2; and (3) performing steps 1 and 2.

[0040] QoE modeling methods based on offline data analysis mainly build QoE models by analyzing large-scale data sets collected offline. These models are usually designed to approximate or predict the user experience quality (QoE) through measurable quality of service (QoS) indicators (such as bit rate, pause time, smoothness, etc.). Usually a linear weighted model (expressed as QoE lin ) or a neural network-based mean opinion score (MOS) model (expressed as QoE nn) as the optimization target. These models do not rely on explicit user feedback during playback, and they usually represent the average user experience. Specifically, this type of technical solution establishes a mathematical relationship between QoS indicators and QoE (usually embodied as MOS or similar aggregate indicators) by analyzing historical playback data (usually does not include real-time feedback for specific users). For example, a function is learned through regression analysis or training a neural network, which inputs a set of QoS indicator values ​​and outputs a predicted average QoE value. These models are then used to guide the decision-making of the ABR algorithm in order to maximize the predicted average QoE value. However, the unified QoE obtained based on offline data analysis is not suitable for the needs of the user. lin or QoE nn The model essentially represents the experience of the "average user" and cannot effectively capture and adapt to the significant differences in QoS sensitivity among different users. Although the model may be trained on user data, its optimization goals and application methods are usually system-level, aiming to improve the overall average experience rather than optimizing the personalized experience of each user.

[0041] Personalized QoE modeling methods based on laboratory research explore and build user-level personalized QoE models by conducting subjective evaluation experiments in a controlled laboratory environment. In these experiments, users are typically recruited to watch a preset video sequence and asked to provide explicit feedback (e.g., ratings) about their viewing experience. This feedback is then used to build a personalized QoE model for each user participating in the experiment. This type of method can achieve user-level modeling and does not necessarily have a direct negative impact on the user experience when collecting feedback (e.g., feedback is collected after the viewing session ends). Specifically, researchers design video clips containing different QoS levels (e.g., different bitrates, freeze modes), have participants watch them in a laboratory environment, and use standardized questionnaires or rating systems (e.g., mean opinion scores) to collect their subjective experience evaluations of each clip. Based on the collected individual user data, a unique QoE model can be fitted for each user, which reflects the user's specific preferences and tolerance for different QoS factors. However, the process of obtaining user feedback (even after a session) can be cumbersome. If it needs to be continued in actual applications, it will significantly affect the smoothness of the user experience. In large-scale production environments, methods that rely on laboratory environments and recruiting users for subjective evaluation are costly and difficult to promote and sustain, which limits their scalability and practicality in large-scale deployments. Laboratory research is usually a one-time data collection and model building, which is difficult to adapt to the dynamic changes in user preferences that may occur over time and lacks the ability to continuously optimize in real-world playback scenarios.

[0042] Personalized QoE analysis methods based on playback intervention attempt to quickly analyze the characteristics or QoS sensitivity of new users by proactively introducing interventions during their actual playback (e.g., temporarily limiting available bandwidth). By observing how users react to these controlled interventions (e.g., whether they exit the video, changes in playback behavior, etc.), the system attempts to build a preliminary personalized model for the user. This type of method operates at the user level and enables continuous (albeit intervention-based) optimization attempts. Specifically, the system might intentionally reduce network bandwidth or introduce simulated freezes at specific moments while the user is watching a video, then monitor changes in user behavior. For example, if a user exits the video even after a slight bandwidth restriction, the system may determine that the user is sensitive to network fluctuations. This information is used to adjust subsequent playback strategies to improve the user's experience. However, proactively introduced playback interventions, such as bandwidth throttling or artificial freezes, directly damage the viewing experience of current users and may lead to user dissatisfaction or churn. Such methods rely on actively changing users' playback conditions to obtain information, rather than learning their preferences by observing their behavior under natural playback conditions. If such interventions are frequently used in large-scale user groups, they may have a negative impact on the platform's overall user satisfaction and retention.

[0043] This disclosure takes into account that some QoE personalization methods mainly target higher-level components in the video streaming architecture, or rely on small-scale laboratory experiments and clear user feedback, which have scalability and practicality issues in actual large-scale deployments. In addition, methods that quickly analyze new users through playback intervention (such as bandwidth restrictions) inevitably reduce the user experience. This disclosure proposes a method for user-level personalized adaptive video streaming that can be deployed on a large scale. It perceives user QoE by online analyzing the current user's natural playback behavior during playback (especially the reaction to freezes), and dynamically adjusts the key parameters of the ABR algorithm accordingly to achieve the optimization goal of dynamically adjusting the ABR algorithm. The following will be combined with Figures 1 to 7 Provide detailed explanation.

[0044] Figure 1 A flowchart illustrating a method for adjusting ABR parameters according to an exemplary embodiment of the present disclosure is shown.

[0045] Reference Figure 1 In step S101, the user's current video playback feature information is obtained.

[0046] As an exemplary embodiment, step S101 is executed in response to the occurrence of a freeze condition satisfying a preset condition. In other words, the occurrence of a freeze condition satisfying the preset condition may trigger the execution of the ABR parameter adjustment method according to an exemplary embodiment of the present disclosure. As an example, the preset condition may include: the number of freezes within a preset duration reaches a preset number.

[0047] Video playback feature information includes but is not limited to: network status information (e.g., network status information in the recent period), player status information (e.g., current player status information), and user playback behavior information (e.g., user playback behavior information in the recent period).

[0048] As an example, the network status information may include, but is not limited to: bandwidth distribution.

[0049] As an example, the player status information may include, but is not limited to: bit rate, buffer status information.

[0050] For example, user playback behavior information may include, but is not limited to, information about freeze reactions (i.e., user reactions to freezes, such as whether the user exited the video due to freezes, etc.) and may further include user viewing time information.

[0051] In addition, the video playback feature information may further include other feature information, such as scene complexity, motion intensity, and other video content-related feature information, which is not limited in the present disclosure.

[0052] In step S102, based on the video playback feature information, the optimal parameter value of the specific ABR parameter of the player is determined.

[0053] Specific ABR parameters include, but are not limited to, parameters used in the ABR algorithm to characterize a user's tolerance for stuttering. Preferably, as an exemplary embodiment, specific ABR parameters include, but are not limited to, parameters used in the ABR algorithm to characterize a user's tolerance for stuttering, parameters used in the ABR algorithm to characterize a user's preference for smoothness, and parameters used in the ABR algorithm to characterize a user's preference for video quality.

[0054] As an exemplary embodiment, step S102 may include determining optimal values ​​for specific ABR parameters using an optimization algorithm based on video playback feature information, wherein the optimization algorithm aims to find parameter values ​​that minimize the probability of users exiting the video. By way of example, the optimization algorithm may be an online Bayesian optimization algorithm, or another online black-box optimization algorithm or reinforcement learning algorithm, which is not limited in this disclosure.

[0055] As an exemplary embodiment, in the step of determining the optimal parameter value of a specific ABR parameter through an optimization algorithm, the video exit probability corresponding to the candidate parameter value can be determined in the following manner: based on the video playback feature information, simulate a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value; then, the sum of the video exit probabilities corresponding to each of these series of video playback trajectories is used as the video exit probability corresponding to the candidate parameter value. Figure 2A detailed description is given here.

[0056] It should be understood that the series of video playback tracks includes: multiple video playback tracks, wherein a video playback track is a virtual playback track for a video, specifically including a virtual playback track of at least one video segment of the video.

[0057] Candidate parameter values ​​are parameter values ​​of specific ABR parameters explored during the search for optimal parameter values. It should be understood that when the specific ABR parameters include: a parameter representing a user's tolerance for stuttering, a parameter representing a user's preference for smoothness, and a parameter representing a user's preference for video quality, the candidate parameter values ​​specifically include: a candidate parameter value for the parameter representing the user's tolerance for stuttering, a candidate parameter value for the parameter representing the user's preference for smoothness, and a candidate parameter value for the parameter representing the user's preference for video quality.

[0058] In step S103 , the parameter value of the specific ABR parameter is updated to the optimal parameter value, so as to select the bit rate of the video stream using the ABR algorithm after the parameter value is updated.

[0059] Specific ABR parameters are used to determine the ABR algorithm's optimization target. Dynamically updating the values ​​of specific ABR parameters is essentially updating the algorithm's optimization target. The ABR algorithm selects an appropriate video stream bitrate based on its optimization target. In other words, the optimization target guides the algorithm's decision-making. Specifically, the ABR algorithm uses specific ABR parameters to determine a user's current tolerance for stuttering, preference for smoothness, and preference for video quality, and selects a bitrate accordingly to improve the user's experience.

[0060] As an example, specific ABR parameters may be parameters in a QoE model, the input of which may be QoS indicator values ​​(e.g., freeze time, video quality, smoothness, etc.), and the output of which is a predicted QoE value. The model is used to guide the decision-making of the ABR algorithm in order to maximize the predicted QoE value.

[0061] Figure 2 A flowchart illustrating a method for simulating a series of possible video playback trajectories starting from a current state of a player according to an exemplary embodiment of the present disclosure is shown.

[0062] Reference Figure 2 In step S201, the current bandwidth value is determined based on the bandwidth distribution.

[0063] As an example, the bandwidth distribution in the video playback feature information may be modeled as a normal distribution, and the current bandwidth value may be obtained by sampling based on the normal distribution.

[0064] In step S202, based on the current bandwidth value and the current buffer status information, a current bit rate selected by the ABR algorithm when a specific ABR parameter is a candidate parameter value is obtained.

[0065] The initial value of the current buffer state information is the buffer state information in the video playback feature information, and the current buffer state information will be updated in real time as the simulation process proceeds.

[0066] In step S203 , a virtual playback track of the j-th video segment of the i-th video is simulated according to the current bit rate, where the initial values ​​of i and j are 1.

[0067] In step S204, based on the freeze time, video quality, smoothness, and historical freeze reaction information corresponding to the virtual playback trajectory of the jth video segment of the ith video, the exit probability of the j+1th video segment of the ith video is predicted, and based on the exit probability, it is determined whether to continue playing the j+1th video segment of the ith video.

[0068] As an example, if the exit probability of the j+1th video segment of the i-th video is higher than a preset probability threshold, it is determined not to continue playing the j+1th video segment of the i-th video.

[0069] The following will be combined Figure 3 Step S204 is described in detail and will not be expanded here.

[0070] In step S205 , if it is determined to continue playing the j+1th video segment of the i-th video, then j=j+1 and the process returns to step S201 to simulate the virtual playback trajectory of the next video segment.

[0071] In step S206, if it is determined that the j+1th video segment of the i-th video will not continue to be played, the exit probability is used as the video exit probability corresponding to the i-th video playback track, and after setting i=i+1 and j=1, the process returns to step S201 to start simulating the virtual playback track for the next video.

[0072] In addition, it should be understood that when the cumulative total duration of the virtual playback trajectory simulated for the current candidate parameter value has reached the set duration, the simulation for the current candidate parameter value is stopped, and the sum of the video exit probabilities corresponding to each simulated video playback trajectory is used as the video exit probability corresponding to the current candidate parameter value.

[0073] Figure 3 A flowchart illustrating a method for predicting an exit probability of a j+1th video segment of an i-th video according to an exemplary embodiment of the present disclosure is shown.

[0074] Reference Figure 3In step S301, based on the current bit rate, the current bandwidth value, the freeze time corresponding to the virtual playback trajectory of the jth video segment of the i-th video, and the historical freeze reaction information, the first exit probability of the j+1th video segment of the i-th video is predicted.

[0075] It should be understood that steps S301-S304 are only executed when the freeze time corresponding to the virtual playback track of the j-th video segment of the i-th video is not 0, otherwise only steps S302-S304 are executed without step S301.

[0076] As an exemplary embodiment, step S301 may include: inputting the time series data into a pre-trained neural network model to obtain a first exit probability output by the neural network model.

[0077] The time series data includes time series data of the following dimensions: bit rate, bandwidth value, freeze time, interval from the last freeze, and interval from the last freeze that caused user exit.

[0078] Figure 5 A schematic diagram illustrating a neural network model for predicting a first exit probability according to an exemplary embodiment of the present disclosure is shown.

[0079] like Figure 5 As shown, the input of the neural network model may include: time series data of five dimensions (each row represents one dimension), representing short-term playback information and long-term user participation information, and the length of the time series data of each dimension is 8 (each column corresponds to a freeze event). For example, the first column corresponds to the freeze event in the virtual playback track of the jth video clip of the i-th video. The first column of data specifically includes: the current bit rate, the current bandwidth value, the freeze time, the interval from the last freeze, and the interval from the last freeze that caused the user to exit. The other columns except the first column correspond to historical freeze events. Specifically, the second column corresponds to the previous freeze event, the third column corresponds to the previous freeze event, and so on. It will not be repeated here. In addition, it should be understood that when the freeze event occurs for the first time when simulating the current candidate parameter value, the historical freeze events corresponding to the other columns except the first column are: the actual freeze event (non-simulated) recorded in the freeze reaction information included in the video playback feature information. Accordingly, the other columns of data are generated based on the freeze reaction information included in the video playback feature information. When a freeze event occurs for the second time when simulating the current candidate parameter value, the first column corresponds to the second freeze event during the simulation, and the second column corresponds to the first freeze event during the simulation. The historical freeze events corresponding to the other columns except the first and second columns are: the actual freeze events (non-simulated) recorded in the freeze reaction information included in the video playback feature information, and so on. No further details will be given.

[0080] The output of the neural network model can be a two-dimensional vector representing the probability of continuing to watch the video and quitting.

[0081] The neural network model can use five 1D convolutional layers to extract features, merge them, and feed them into a 64D fully connected layer. This layer then passes through a 2D fully connected layer, and finally outputs the exit probability using a softmax activation function. Alternatively, the model can use other types of network structures, such as recurrent neural networks (RNNs) or long short-term memory networks (LSTMs), to better capture time series dependencies.

[0082] This neural network model can use the cross entropy loss function.

[0083] In step S302 , based on the smoothness corresponding to the virtual playback trajectory of the jth video segment of the ith video, a second exit probability of the j+1th video segment of the ith video is predicted.

[0084] As an exemplary embodiment, the smoothness corresponding to the virtual playback trajectory of the j-th video segment of the ith video may be input into the first statistical model to obtain a second exit probability output by the first statistical model.

[0085] As an example, the first statistical model may be a model obtained by performing statistics on the collected relevant data.

[0086] In step S303 , based on the video quality corresponding to the virtual playback track of the jth video segment of the ith video, a third exit probability of the j+1th video segment of the ith video is predicted.

[0087] As an exemplary embodiment, the smoothness corresponding to the virtual playback trajectory of the j-th video segment of the i-th video may be input into the second statistical model to obtain a third exit probability output by the second statistical model.

[0088] As an example, the second statistical model may be a model obtained by performing statistics on the collected relevant data.

[0089] In step S304, the exit probability of the j+1th video segment of the ith video is determined based on the first exit probability, the second exit probability, and the third exit probability.

[0090] As an exemplary embodiment, step S304 may include: taking the sum of the first exit probability, the second exit probability, and the third exit probability as the exit probability of the j+1th video segment of the i-th video.

[0091] Figure 4 A schematic diagram illustrating an ABR parameter adjustment system according to an exemplary embodiment of the present disclosure.

[0092] Reference Figure 4The system leverages production data, combined with playback logs and user engagement metrics, to dynamically optimize the ABR algorithm's objectives. It primarily includes an online Bayesian optimization module, a Monte Carlo sampling module, and a hybrid exit probability predictor.

[0093] The core ideas and mechanisms of this system mainly include: optimization based on user participation, personalized lag perception modeling, and dynamic target adjustment.

[0094] Specifically, this paper uses user dropout probability at the video clip level as a key metric, linking optimization objectives with user engagement. By analyzing the relationship between dropout probability and QoS metrics (video quality, smoothness, and freeze time), it was found that freeze time has a far greater impact on dropout probability than other factors.

[0095] Personalized jamming perception modeling: In view of the significant individual differences and certain temporal dynamics in users' tolerance for jamming, this disclosure establishes a personalized model for jamming events, while a general statistical model is used for video quality and smoothness that have less impact.

[0096] Dynamic target adjustment: This disclosure supports any ABR algorithm and modifies its key parameters at runtime to modify the optimization target through a dynamic QoE adjustment module.

[0097] Online Bayesian Optimization: This paper considers the optimal parameter value for each user's QoE parameter (i.e., a specific ABR parameter, for example, a parameter that affects the ABR algorithm's tolerance to lag) as an online black-box optimization problem. The goal is to find the parameter value x* that minimizes the user's overall predicted exit probability, where x*∈argmin x∈X f Rexit (x, N, S), where x represents the QoE parameter, N represents the bandwidth distribution, and S represents the current state (player and user state). Online Bayesian optimization is used to independently explore and determine candidate ABR parameter values ​​for each user. It uses historical trial data to fit a proxy model (such as a Gaussian process) and iteratively optimizes the objective function by determining the next QoE parameter value x to explore by maximizing an acquisition function (such as the probability of improvement). When the QoE adjustment mechanism is triggered (for example, after the user experiences a certain amount of lag), the optimization process uses the previously optimized configuration as the initialization point for subsequent iterations.

[0098] Monte Carlo sampling and simulation environment,In order to evaluate the effect of specific QoE parameter value settings and consider the long-term impact of the current bitrate selection on immediate and future QoE, this disclosure uses Monte Carlo sampling to generate playback trajectories in a simulation environment. The simulator takes into account the historical bandwidth distribution (modeled as a normal distribution) and uses candidate QoE parameter values ​​to adjust the ABR algorithm for virtual playback starting from the current playback state. The simulated playback duration and user exit behavior refer to the actual viewing patterns of online users. In this way, a series of possible playback trajectories and QoS sequences can be generated for a given QoE parameter value.

[0099] Exit Rate Predictor: This paper develops a hybrid exit probability prediction model to quantify the user experience under different parameter settings. The model combines personalized neural network modeling (for freeze time) and holistic statistical modeling (for video quality and smoothness indicators).

[0100] According to an exemplary embodiment of the present disclosure, a dynamic optimization target adjustment based on implicit user feedback (exit probability) is proposed: instead of relying on explicit user ratings, the user QoE is perceived by analyzing the user's natural exit behavior during playback (especially the reaction to lag), and the optimization target of the ABR algorithm is dynamically adjusted accordingly, which makes personalized optimization non-intrusive to users.

[0101] According to exemplary embodiments of the present disclosure, a hybrid personalized dropout rate prediction model is proposed. This model addresses the varying degrees of impact of different QoS factors on QoE, using a hybrid modeling approach. For the most influential lag events, a personalized neural network model is used to capture complex nonlinear relationships and individual differences; for less influential factors such as video quality and smoothness, a holistic statistical model is used. This approach balances accuracy and efficiency.

[0102] According to exemplary embodiments of this disclosure, a parameter optimization mechanism combining Monte Carlo sampling and online Bayesian optimization is proposed. Monte Carlo sampling is used to generate QoS trajectories for given parameter values ​​in a simulation environment for evaluating the exit probability predictor. This is then combined with online Bayesian optimization for continuous parameter value exploration and optimization. This approach can handle the uncertainty of network and user behavior and adapt to dynamic changes in user perception.

[0103] According to exemplary embodiments of the present disclosure, a system and method for personalized optimization of adaptive video streams for user-level quality of experience (QoE) are provided. These methods implement user-level personalized QoE optimization: dynamic and personalized adjustment of the ABR algorithm optimization objectives is achieved based on the differences in QoS perception and tolerance among different users. Seamless integration and low interference provide a personalized optimization solution that can be seamlessly integrated with existing ABR systems and does not rely on explicit user feedback during the optimization process, thereby not negatively interfering with the user experience. Dynamic adaptability captures individual differences and temporal dynamics in user QoE perception, enabling continuous parameter optimization and adjustment.

[0104] The present disclosure can bring the following beneficial effects:

[0105] Significantly improve total user viewing time: In a large-scale A / B test, this disclosure increased total viewing time by 0.15% compared to the existing production algorithm;

[0106] Improved QoS indicators: achieved a 0.1% increase in average bit rate and a 1.3% reduction in total stall time;

[0107] Superior performance under bandwidth-constrained conditions: Under bandwidth conditions below 2000kbps, the delay time is reduced by about 15% while maintaining comparable video quality.

[0108] Achieve effective user-level adaptation: Dynamically adjust algorithm parameters based on the individual user's sensitivity to lag, and adopt a more conservative strategy for users with low tolerance for lag.

[0109] Figure 6 A structural block diagram of an ABR parameter adjustment device according to an exemplary embodiment of the present disclosure is shown.

[0110] Reference Figure 6 According to an exemplary embodiment of the present disclosure, an ABR parameter adjustment device 600 includes: a feature information acquisition unit 601, an optimization unit 602, and a parameter value updating unit 603.

[0111] Specifically, the feature information acquisition unit 601 is configured to acquire the user's current video playback feature information, wherein the video playback feature information includes: network status information, player status information, and user playback behavior information.

[0112] The optimization unit 602 is configured to determine the optimal parameter value of the specific ABR parameter of the player based on the video playback feature information, wherein the specific ABR parameter includes: a parameter in the ABR algorithm used to characterize the user's tolerance to lag.

[0113] The parameter value updating unit 603 is configured to update the parameter value of the specific ABR parameter to the optimal parameter value, so as to select the bit rate of the video stream using the ABR algorithm after the parameter value is updated.

[0114] As an exemplary embodiment, the specific ABR parameters may further include: parameters in the ABR algorithm used to characterize the user's preference for smoothness, and / or parameters in the ABR algorithm used to characterize the user's preference for video quality.

[0115] As an exemplary embodiment, the network status information may include: bandwidth distribution; the player status information may include: bit rate, buffer status information; the user playback behavior information may include: freeze response information.

[0116] As an exemplary embodiment, the optimization unit 602 can be configured to: determine the optimal parameter value of the specific ABR parameter through an optimization algorithm based on the video playback feature information, wherein the goal of the optimization algorithm is to find a parameter value that can minimize the user's video exit probability.

[0117] As an exemplary embodiment, the optimization unit 602 can be configured to determine the video exit probability corresponding to the candidate parameter value in the following manner: based on the video playback feature information, simulate a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value; and take the sum of the video exit probabilities corresponding to each of the series of video playback trajectories as the video exit probability corresponding to the candidate parameter value.

[0118] As an exemplary embodiment, the optimization unit 602 may be configured to: determine a current bandwidth value based on a bandwidth distribution; obtain a current bit rate selected by the ABR algorithm when the specific ABR parameter is the candidate parameter value based on the current bandwidth value and current buffer status information; simulate a virtual playback trajectory of the jth video segment of the ith video according to the current bit rate, where the initial values ​​of i and j are 1; predict an exit probability of the j+1th video segment of the ith video based on the freeze time, video quality, smoothness, and historical freeze response information corresponding to the virtual playback trajectory of the jth video segment of the ith video, and determine whether to continue playing the j+1th video segment of the ith video based on the exit probability; if it is determined to continue playing the j+1th video segment of the ith video, set j=j+1 and return to executing the determination of the current bandwidth value based on the bandwidth distribution; if it is determined not to continue playing the j+1th video segment of the ith video, use the exit probability as the video exit probability corresponding to the ith video playback trajectory, set i=i+1 and j=1 and return to executing the determination of the current bandwidth value based on the bandwidth distribution.

[0119] As an exemplary embodiment, the optimization unit 602 can be configured to: predict a first exit probability of the j+1th video segment of the ith video based on the current bit rate, the current bandwidth value, the freeze time corresponding to the virtual playback trajectory of the jth video segment of the ith video, and the historical freeze reaction information; predict a second exit probability of the j+1th video segment of the ith video based on the smoothness corresponding to the virtual playback trajectory of the jth video segment of the ith video; predict a third exit probability of the j+1th video segment of the ith video based on the video quality corresponding to the virtual playback trajectory of the jth video segment of the ith video; and determine the exit probability of the j+1th video segment of the ith video based on the first exit probability, the second exit probability, and the third exit probability.

[0120] As an exemplary embodiment, the optimization unit 602 can be configured to: input the time series data into a pre-trained neural network model to obtain a first exit probability output by the neural network model; wherein the time series data includes time series data of the following dimensions: bit rate, bandwidth value, freeze time, interval from the last freeze, and interval from the last freeze that caused the user to exit.

[0121] As an exemplary embodiment, the feature information acquisition unit 601 may be configured to: in response to a freeze condition occurring that satisfies a preset condition, acquire feature information of the user's current video playback.

[0122] Regarding the ABR parameter adjustment device 600 in the above embodiment, the specific manner in which each unit performs operations has been described in detail in the embodiment of the method and will not be elaborated on here.

[0123] Furthermore, it should be understood that the various units in the ABR parameter adjustment apparatus 600 according to the exemplary embodiment of the present disclosure may be implemented as hardware components and / or software components. Those skilled in the art may implement the various units using, for example, a field programmable gate array (FPGA) or an application-specific integrated circuit (ASIC), depending on the processing performed by the defined units.

[0124] Figure 7 FIG. 1 shows a structural block diagram of an electronic device according to an exemplary embodiment of the present disclosure. Figure 7 The electronic device 700 includes: at least one memory 701 and at least one processor 702, wherein the at least one memory 701 stores a set of computer-executable instructions. When the computer-executable instruction set is executed by the at least one processor 702, the ABR parameter adjustment method as described in the above exemplary embodiment is executed.

[0125] As an example, the electronic device 700 may be a PC, a tablet device, a personal digital assistant, a smart phone, or other device capable of executing the above-mentioned instruction set. Here, the electronic device 700 is not necessarily a single electronic device, but may also be any device or circuit collection capable of executing the above-mentioned instructions (or instruction set) individually or in combination. The electronic device 700 may also be part of an integrated control system or system manager, or may be configured as a portable electronic device interconnected with a local or remote (e.g., via wireless transmission) interface.

[0126] In electronic device 700, processor 702 may include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, processor 702 may also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, etc.

[0127] The processor 702 can execute instructions or codes stored in the memory 701, wherein the memory 701 can also store data. Instructions and data can also be sent and received over the network via the network interface device, wherein the network interface device can use any known transmission protocol.

[0128] The memory 701 may be integrated with the processor 702, for example, by placing RAM or flash memory within an integrated circuit microprocessor or the like. Furthermore, the memory 701 may comprise a separate device, such as an external disk drive, a storage array, or any other storage device usable by a database system. The memory 701 and the processor 702 may be operatively coupled or may communicate with each other, for example, via an I / O port, a network connection, or the like, such that the processor 702 can access files stored in the memory.

[0129] In addition, the electronic device 700 may further include a multimedia resource display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, a mouse, a touch input device, etc.) All components of the electronic device 700 may be connected to each other via a bus and / or a network.

[0130] According to an exemplary embodiment of the present disclosure, a computer-readable storage medium storing instructions may also be provided, wherein when the instructions are executed by at least one processor, the at least one processor is prompted to perform the ABR parameter adjustment method as described in the above exemplary embodiment. Examples of computer-readable storage media here include: read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, Blu-ray or optical disk storage, hard disk drive (HDD), solid state drive (SSD), card storage (such as, multimedia card, secure digital (SD) card or ultra-fast digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk and any other device, any other device configured to store the computer program and any associated data, data files and data structures in a non-transitory manner and provide the computer program and any associated data, data files and data structures to a processor or computer so that the processor or computer can execute the computer program. The computer program in the above-mentioned computer-readable storage medium can be run in an environment deployed in a computer device such as a client, a host, an agent device, a server, etc. In addition, in one example, the computer program and any associated data, data files and data structures are distributed on a networked computer system so that the computer program and any associated data, data files and data structures are stored, accessed and executed in a distributed manner by one or more processors or computers.

[0131] According to an exemplary embodiment of the present disclosure, a computer program product may also be provided. Instructions in the computer program product may be executed by at least one processor to implement the method for adjusting the ABR parameters as described in the above exemplary embodiment.

[0132] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0133] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for adjusting ABR parameters, characterized in that: include: Obtaining the user's current video playback feature information, wherein the video playback feature information includes: network status information, player status information, and user playback behavior information; Determining optimal parameter values ​​of specific ABR parameters of the player based on the video playback characteristic information, wherein the specific ABR parameters include: parameters in the ABR algorithm used to characterize the user's tolerance to lag; The parameter value of the specific ABR parameter is updated to the optimal parameter value, so as to select the bit rate of the video stream using the ABR algorithm after the parameter value is updated.

2. The adjustment method according to claim 1, characterized in that: The specific ABR parameters also include: parameters in the ABR algorithm used to characterize the user's preference for smoothness, and / or parameters in the ABR algorithm used to characterize the user's preference for video quality.

3. The adjustment method according to claim 1, characterized in that: The network status information includes: bandwidth distribution; The player status information includes: bit rate, buffer status information; The user playback behavior information includes: freeze reaction information.

4. The adjustment method according to claim 1, characterized in that: The determining, based on the video playback feature information, an optimal parameter value of a specific ABR parameter of the player includes: Based on the video playback feature information, an optimal parameter value of the specific ABR parameter is determined through an optimization algorithm, wherein a goal of the optimization algorithm is to find a parameter value that can minimize the probability of a user exiting the video.

5. The adjustment method according to claim 4, characterized in that: In the step of determining the optimal parameter value of the specific ABR parameter by an optimization algorithm based on the video playback feature information, the video exit probability corresponding to the candidate parameter value is determined by the following method: Based on the video playback feature information, simulating a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value; The sum of the video exit probabilities corresponding to the series of video playback trajectories is used as the video exit probability corresponding to the candidate parameter value.

6. The adjustment method according to claim 5, characterized in that: The simulating, based on the video playback feature information, a series of possible video playback trajectories starting from the current state of the player when the specific ABR parameter is the candidate parameter value, includes: determining a current bandwidth value based on the bandwidth distribution; obtaining, based on a current bandwidth value and current buffer status information, a current bit rate selected by the ABR algorithm when the specific ABR parameter is the candidate parameter value; Simulate the virtual playback trajectory of the jth video segment of the i-th video according to the current bit rate, where the initial values ​​of i and j are 1; Based on the freeze time, video quality, smoothness, and historical freeze reaction information corresponding to the virtual playback trajectory of the j-th video segment of the i-th video, predict the exit probability of the j+1-th video segment of the i-th video, and determine whether to continue playing the j+1-th video segment of the i-th video based on the exit probability; If it is determined to continue playing the j+1th video segment of the i-th video, then set j=j+1 and return to the step of determining the current bandwidth value based on the bandwidth distribution; If it is determined not to continue playing the j+1th video segment of the i-th video, the exit probability is used as the video exit probability corresponding to the i-th video playback track, and after setting i=i+1 and j=1, the process returns to determine the current bandwidth value based on bandwidth distribution.

7. The adjustment method according to claim 6, characterized in that: The predicting the exit probability of the j+1th video segment of the i-th video includes: Predict the first exit probability of the j+1th video segment of the i-th video based on the current bit rate, the current bandwidth value, the freeze time corresponding to the virtual playback trajectory of the j-th video segment of the i-th video, and the historical freeze reaction information; Predicting the second exit probability of the j+1th video segment of the i-th video based on the smoothness corresponding to the virtual playback trajectory of the j-th video segment of the i-th video; Predicting the third exit probability of the j+1th video segment of the i-th video based on the video quality corresponding to the virtual playback trajectory of the j-th video segment of the i-th video; Based on the first exit probability, the second exit probability, and the third exit probability, the exit probability of the j+1th video segment of the ith video is determined.

8. The adjustment method according to claim 7, characterized in that: The predicting a first exit probability of the j+1th video segment of the i-th video includes: Inputting the time series data into a pre-trained neural network model to obtain a first exit probability output by the neural network model; The time series data includes the following dimensions: bit rate, bandwidth value, freeze time, interval from the last freeze, and interval from the last freeze that caused user exit.

9. The adjustment method according to claim 1, characterized in that: The step of obtaining the user's current video playback feature information includes: In response to the occurrence of a freeze condition meeting a preset condition, the user's current video playback feature information is obtained.

10. An ABR parameter adjustment device, characterized in that: include: The feature information acquisition unit is configured to acquire the user's current video playback feature information, wherein the video playback feature information includes: network status information, player status information, and user playback behavior information; An optimization unit is configured to determine an optimal parameter value of a specific ABR parameter of the player based on the video playback characteristic information, wherein the specific ABR parameter includes: a parameter in the ABR algorithm used to characterize the user's tolerance to lag; The parameter value updating unit is configured to update the parameter value of the specific ABR parameter to the optimal parameter value, so as to select the bit rate of the video stream using the ABR algorithm after the parameter value is updated.

11. An electronic device, characterized in that: include: at least one processor; at least one memory storing computer-executable instructions, When the computer-executable instructions are executed by the at least one processor, the computer-executable instructions prompt the at least one processor to perform the ABR parameter adjustment method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to perform the method for adjusting the ABR parameters according to any one of claims 1 to 9.

13. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by at least one processor, the method for adjusting the ABR parameters according to any one of claims 1 to 9 is implemented.

Images