Video user preference prediction method and device in dynamic vehicle-mounted environment and electronic equipment

Abstract

The application provides a video user preference prediction method and device in a dynamic vehicle environment and electronic equipment, wherein after the influence factor of quality of experience (QoE) of a next video slice to be played is adjusted by using the wireless channel parameter of a current played video slice, the first QoE score is obtained by combining a target QoE scoring model and the wireless channel parameter; all adjusted QoE influence factors are traversed, and each time, only a preset disturbance amount is superimposed on one QoE influence factor, and other QoE influence factors remain unchanged, and then the target QoE scoring model and the wireless channel parameter are combined for prediction, and finally, multiple second QoE scores are obtained; after the target sensitivity coefficients between the multiple second QoE scores and the first QoE score are calculated, weight distribution is performed, and the preference prediction result of the first QoE influence factor of the user is obtained. The application can accurately obtain the prediction preference of the user for various QoE influence factors in the dynamic vehicle environment.

Description

Technical Field

[0001] This invention belongs to the field of vehicle networking technology, specifically relating to a method for predicting video user preferences in a dynamic in-vehicle environment. Background Technology

[0002] With the continuous improvement of vehicle intelligence and connectivity, vehicles are gradually acquiring the capabilities of environmental perception, status analysis, decision control, and service execution. They can act as intelligent agents for vehicle users, providing on-demand video streaming services. During in-vehicle video streaming services, different users exhibit varying degrees of concern regarding Quality of Experience (QoE) factors such as video resolution, frame rate, total buffering time, and initial buffering time. Therefore, accurately acquiring user preferences for various QoE influencing factors becomes a crucial foundation for achieving personalized video services and improving service quality assurance capabilities.

[0003] In related technologies, on the one hand, fixed weights, fixed parameters, or preset user types are used to describe the degree of user attention to QoE influencing factors such as resolution, frame rate, total stuttering time, and initial buffering time. However, the modeling basis of this method is mostly based on static subjective rating data or predetermined user needs, lacking the ability to dynamically update user preferences according to environmental changes. Therefore, it is difficult to accurately represent the real-time changes in user preferences in the context of vehicle-to-everything (V2X) video streaming services. On the other hand, video bitrate, resolution, frame rate, and buffering strategies are dynamically adjusted according to real-time channel status, bandwidth changes, or link quality changes to balance video quality, playback smoothness, and transmission efficiency. However, in real-world V2X video streaming service scenarios, high-speed vehicle movement, road scene switching, and wireless link fluctuations all lead to continuous changes in channel status, which in turn cause dynamic changes in QoE influencing factors such as video resolution, frame rate, total stuttering time, and initial buffering time. Therefore, this method lacks an organic connection between changes in environmental status, video parameters, and user preferences, resulting in an inability to accurately predict user preferences in real-time in V2X video streaming service scenarios.

[0004] Therefore, how to accurately obtain users' predictive preferences for various QoE influencing factors in a dynamic in-vehicle environment has become an urgent problem to be solved. Summary of the Invention

[0005] To address the aforementioned problems in existing technologies, this invention provides a method for predicting video user preferences in dynamic vehicular environments, solving the problem that related technologies cannot accurately obtain user preferences for various QoE influencing factors in dynamic vehicular environments.

[0006] According to a first aspect of the present invention, a method for predicting video user preferences in a dynamic vehicular environment is provided, comprising: The first QoE impact factor of the next video slice to be played is adaptively adjusted using the wireless channel parameters of the currently playing video slice to obtain the second QoE impact factor. The target QoE scoring model is used to predict the second QoE influence factor and wireless channel parameters to obtain the first QoE score of the next video slice to be played. The second QoE influence factor is iterated through, and a preset perturbation is superimposed on the current second QoE influence factor in each round to obtain the third QoE influence factor; the third QoE influence factor, the remaining second QoE influence factors, and the wireless channel parameters are predicted using the target QoE scoring model to obtain the second QoE score. After the iteration is completed, multiple second QoE scores are obtained. The target sensitivity coefficients between multiple second QoE scores and first QoE scores are calculated separately, and weights are assigned based on the target sensitivity coefficients to obtain the user's preference prediction results for the first QoE influence factor.

[0007] In one embodiment of the present invention, the wireless channel state parameters include effective bandwidth and effective signal-to-noise ratio; The process of adaptively adjusting the first QoE impact factor of the next video slice to be played using the wireless channel parameters of the currently playing video slice to obtain the second QoE impact factor includes: Obtain the original bitrate of the next video slice to be played; Substituting the effective bandwidth, effective signal-to-noise ratio, and original coding rate into the formula for actual usable transmission coding rate, we obtain the actual usable transmission coding rate. The first QoE influence factor is adaptively adjusted based on the actual available transmission coding rate to obtain the second QoE influence factor; The formula for the actual usable transmission coding rate is as follows: ; In the above formula, The actual usable transmission coding rate, The original encoded bitrate, For efficiency coefficient, For effective bandwidth, For effective signal-to-noise ratio.

[0008] In one embodiment of the present invention, the first QoE influence factor includes the original resolution and the original frame rate; the wireless channel parameters also include effective delay spread; The adaptive adjustment of the first QoE impact factor based on the actual available transmission coding rate to obtain the second QoE impact factor includes: The target resolution is obtained by substituting the actual available transmission coding bitrate, the original resolution, and the original coding bitrate into the resolution adjustment formula; Calculate the product between the preset frame rate protection threshold and the original encoding bitrate; if the actual available transmission encoding bitrate is not less than the product, then use the original frame rate as the target frame rate; if the actual available transmission encoding bitrate is less than the product, then substitute the original frame rate, the preset frame rate protection threshold, and the actual available transmission encoding bitrate into the frame rate adjustment formula to calculate the target frame rate. The resolution adjustment formula is as follows: ; In the above formula, For the target resolution, Original resolution; The frame rate adjustment formula is as follows: ; In the above formula, For the target frame rate, The original frame rate, This is the preset frame rate protection threshold.

[0009] In one embodiment of the present invention, the first QoE impact factor further includes the original total stutter time and the original initial buffer duration; The method of adaptively adjusting the first QoE influence factor based on the actual available transmission coding rate to obtain the second QoE influence factor further includes: Obtain the playback duration of the next video segment to be played; input the effective latency extension, playback duration, and original total stuttering time into the total stuttering time adjustment formula to calculate the target total stuttering time; Obtain the cumulative data amount and latency overhead of the next video slice to be played before playback; and substitute the cumulative data amount, latency overhead and actual available transmission coding bitrate into the initial buffer duration adjustment formula to calculate the target initial buffer duration; The formula for adjusting the total lag time is as follows: ; In the above formula, The target is the total lag time. This represents the original total lag time. This is an empirical coefficient. For playback duration, , This is an empirical coefficient. To adjust the root mean square of the delay spread; The formula for adjusting the initial buffer duration is as follows: ; In the above formula, The target initial buffer duration, L For latency overhead, b This represents the cumulative amount of data.

[0010] In one embodiment of the present invention, the step of calculating the target sensitivity coefficients between a plurality of second QoE scores and a first QoE score includes: The individual differences are obtained by calculating the differences between multiple second QoE scores and first QoE scores; The ratios of individual differences and preset disturbances are calculated to obtain the initial sensitivity coefficients corresponding to each first QoE influence factor; the absolute value of the initial sensitivity coefficients is used as the target sensitivity coefficients.

[0011] In one embodiment of the present invention, weight allocation is performed based on the target sensitivity coefficient to obtain the user's preference prediction result for the first QoE influence factor, including: summing the target sensitivity coefficients to obtain the total sensitivity coefficient; substituting the target sensitivity coefficient and the total sensitivity coefficient into the normalization formula to calculate the user's predicted preference weight for each first QoE influence factor; and summing the predicted preference weights to obtain the preference prediction result. The normalization formula is as follows: ; In the above formula, For the first k The predicted preference weights corresponding to the first QoE impact factor k Index of the first QoE impact factor For the first k The target sensitivity coefficient corresponding to the first QoE impact factor The overall sensitivity coefficient is... K This represents the number of first QoE impact factors.

[0012] In one embodiment of the present invention, the target QoE scoring model is trained through the following process: Under a pre-defined ideal network environment, a first sample dataset is collected; the first sample dataset includes the fourth QoE impact factor and the first QoE score label corresponding to multiple independent video samples; The first QoE scoring model is trained using the first sample dataset to obtain the trained second QoE scoring model. In a dynamic vehicle environment, a second sample dataset is collected; the second sample dataset includes wireless channel parameter samples of the first video slice samples and the fifth QoE influence factor of the second video slice samples; the first video slice samples and the second video slice samples are temporally adjacent video slices; The trained first QoE scoring model is used to predict the second sample dataset to obtain the second QoE scoring label; The second QoE score label is corrected using wireless channel parameter samples to obtain the corrected third QoE score label; The second QoE scoring model was trained using the corrected third QoE score label, wireless channel parameter samples, and the fifth QoE influence factor to obtain the trained target QoE scoring model.

[0013] In one embodiment of the present invention, the wireless channel parameter samples include effective signal-to-noise ratio samples and effective delay spread samples; The step of correcting the second QoE scoring label using wireless channel parameter samples to obtain the corrected third QoE scoring label includes: Substitute the reference signal-to-noise ratio, reference delay spread, effective signal-to-noise ratio sample, and effective delay spread sample into the QoE score correction formula to obtain the QoE score correction value; then sum the QoE score correction value and the second QoE score label to obtain the third QoE score label. The QoE score correction formula is as follows: ; In the above formula, This is the corrected value for the QoE score. and Divided into empirical weighting coefficients, For reference signal-to-noise ratio, For reference delay spread, For effective delay expansion samples, This is a sample with an effective signal-to-noise ratio.

[0014] According to a second aspect of the present invention, a video user preference prediction device for a dynamic in-vehicle environment is provided, the device comprising: The adjustment module is used to adaptively adjust the first QoE impact factor of the next video slice to be played by using the wireless channel parameters of the currently playing video slice, so as to obtain the second QoE impact factor. The prediction module is used to predict the second QoE influence factor and wireless channel parameters using the target QoE scoring model, so as to obtain the first QoE score of the next video slice to be played. The traversal and scoring module is used to traverse the second QoE influence factor, add a preset perturbation amount to the current second QoE influence factor in each round to obtain the third QoE influence factor; use the target QoE scoring model to predict the third QoE influence factor, the remaining second QoE influence factors and wireless channel parameters to obtain the second QoE score. After the traversal is completed, multiple second QoE scores are obtained. The calculation module is used to calculate the target sensitivity coefficients between multiple second QoE scores and first QoE scores, and to perform weight allocation based on the target sensitivity coefficients to obtain the user's preference prediction results for the first QoE influence factor.

[0015] According to a third aspect of the present invention, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, wherein the executable instruction causes the processor to perform an operation corresponding to the method described in the first aspect.

[0016] According to the scheme provided in the embodiments of the present invention, the first QoE influence factor of the next video slice to be played is adaptively adjusted using the wireless channel parameters of the currently playing video slice to obtain the second QoE influence factor; the second QoE influence factor and the wireless channel parameters are predicted using the target QoE scoring model to obtain the first QoE score of the next video slice to be played; the second QoE influence factors are traversed, and a preset perturbation is superimposed on the current second QoE influence factor in each round to obtain the third QoE influence factor; the third QoE influence factor, the remaining second QoE influence factors, and the wireless channel parameters are predicted using the target QoE scoring model to obtain the second QoE score; after traversal, multiple second QoE scores are obtained; the target sensitivity coefficient between the multiple second QoE scores and the first QoE score is calculated respectively, and weight allocation is performed based on the target sensitivity coefficient to obtain the user's preference prediction result for the first QoE influence factor. In this process, environmental parameters of the currently playing video slice and the QoE impact factor of the next video slice to be played are collected. Based on the environmental parameters, the QoE impact factor is adjusted for subsequent preference prediction processing. That is, the final preference weights dynamically change according to environmental influences, enabling dynamic updates of user preferences based on environmental changes and ensuring that the next video slice to be played is adapted to actual network transmission capabilities. This avoids a disconnect between ideal settings and real channel conditions, ensuring that the QoE impact factor prediction and preference results reflect the actual viewing experience available to users, improving the accuracy and effectiveness of strategy decisions. By using a target QoE scoring model to predict based on QoE impact factors and wireless channel parameters, changes in wireless channel parameters, video parameters, and user preferences can be organically correlated. Furthermore, by applying preset perturbations to each QoE impact factor and performing subsequent weight allocation processing, the relative importance of four types of influencing factors—resolution, frame rate, total stuttering time, and initial buffering time—towards the current wireless channel parameters can be quantified. In summary, this method can accurately obtain user preferences for various QoE impact factors in a dynamic vehicular environment.

[0017] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] Figure 1 A flowchart illustrating a method for predicting user preferences for video streaming services in dynamic in-vehicle environments, provided by an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the effect of a vehicle user 1's preference prediction results for different QoE influencing factors in a dynamic in-vehicle environment, as provided in an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the effect of predicting the preferences of vehicle user 2 for different QoE influencing factors in a dynamic in-vehicle environment, as provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of a video streaming service user preference prediction device for dynamic vehicle environments provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention. Based on the examples in the present invention, all other examples obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Figure 1 This is a schematic diagram of a video user preference prediction method in a dynamic vehicle environment provided by an embodiment of the present invention. The video user preference prediction method in a dynamic vehicle environment provided by an embodiment of the present invention can be executed by an electronic device, which can be a server.

[0021] like Figure 1 As shown, a method for predicting user preferences for video streaming services in dynamic in-vehicle environments includes: S101. Adaptively adjust the first QoE impact factor of the next video slice to be played using the wireless channel parameters of the currently playing video slice to obtain the second QoE impact factor.

[0022] Specifically, wireless channel parameters include, but are not limited to, effective bandwidth, effective signal-to-noise ratio (SNR), and effective latency spread. The first QoE impact factor refers to factors affecting the user's experience quality when watching the next video segment to be played, including resolution, frame rate, total stutter time, and initial buffer duration. Resolution and frame rate characterize video quality and temporal continuity, while total stutter time and initial buffer duration are used to characterize the negative impact of playback interruptions and access waiting on QoE. When users are driving and watching videos in various typical vehicle-mounted propagation scenarios such as densely populated urban areas, highways, and suburban roads, the wireless channel parameters of the user watching the currently playing video segment can be collected in real time through the vehicle's built-in wireless communication module. The corresponding resolution and frame rate can be found in the pre-encoding version library using the video content ID and segment number corresponding to the next video segment to be played; the initial buffer duration and the ideal design target value of the total stutter time can be determined by the player's preset strategy, which can be 0. Wireless channel parameters determine network transmission capabilities, so the first QoE impact factor is further adjusted based on effective bandwidth, effective SNR, and effective latency spread to obtain the second QoE impact factor.

[0023] S102. Using the target QoE scoring model, predict the second QoE influence factor and wireless channel parameters to obtain the first QoE score of the next video slice to be played.

[0024] Specifically, the second QoE influence factor and the wireless channel parameters can be fused in a concatenated form to obtain a feature vector. Then, the feature vector is normalized and processed, and the processed feature vector is input into the target QoE scoring model for prediction to obtain the first QoE score for the user watching the next video slice to be played.

[0025] S103. Traverse the second QoE influence factors and add a preset perturbation amount to the current second QoE influence factor in each round to obtain the third QoE influence factor; use the target QoE scoring model to predict the third QoE influence factor, the remaining second QoE influence factors and the wireless channel parameters to obtain the second QoE score. After traversal, multiple second QoE scores are obtained.

[0026] Specifically, the second QoE influence factor consists of adjusted resolution, frame rate, initial buffer duration, and total stutter time. These four adjusted values ​​are iterated over. In each iteration, any one of the adjusted resolution, frame rate, initial buffer duration, or total stutter time is added to a preset perturbation amount. For example, the adjusted resolution and the preset perturbation amount are summed to obtain the superimposed resolution. Then, the superimposed resolution, adjusted frame rate, adjusted initial buffer duration, adjusted total stutter time, and wireless channel parameters are reassembled and normalized. The processed feature vector is then input back into the target QoE scoring model for prediction, resulting in a new QoE score, called the second QoE score. This process of adding any one of the adjusted frame rate, adjusted initial buffer duration, and adjusted total stutter time to the preset perturbation amount continues until each second QoE influence factor corresponds to a final second QoE score.

[0027] S104. Calculate the target sensitivity coefficients between multiple second QoE scores and first QoE scores respectively, and assign weights based on the target sensitivity coefficients to obtain the user's preference prediction results for the first QoE influence factor.

[0028] Specifically, the target sensitivity coefficient reflects the intensity of user experience fluctuations when QoE influencing factors change slightly. After weighting, each of the first QoE influencing factors corresponds to a preference weight. The preference weight indicates which influencing factor in the first QoE influencing factors the user values ​​more when watching the next video segment to be played. After calculating the target sensitivity coefficient between each second QoE score and the first QoE score, the target sensitivity coefficient is normalized for weighting, ultimately yielding the preference prediction result.

[0029] It is understood that in the implementation of this invention, environmental parameters of the currently playing video slice and the QoE impact factor of the next video slice to be played are collected. The QoE impact factor is adjusted based on the environmental parameters and then used for subsequent preference prediction processing. That is, the final preference weights change dynamically due to environmental influences. This achieves the ability to dynamically update user preferences based on environmental changes and ensures that the next video slice to be played adapts to actual network transmission capabilities, avoiding a disconnect between ideal settings and the real channel. This ensures that the QoE impact factor prediction and preference results reflect the actual viewing experience available to the user, improving the accuracy and effectiveness of strategy decisions. By using a target QoE scoring model to predict based on the QoE impact factor and wireless channel parameters, changes in environmental state, video parameters, and user preferences can be organically correlated. Based on this, by applying a preset perturbation to each QoE impact factor and performing subsequent weight allocation processing, the relative importance that users place on four types of influencing factors—resolution, frame rate, total stuttering time, and initial buffering time—under the current wireless channel parameters can be quantified. In summary, this allows for more accurate acquisition of user preferences for various QoE impact factors in dynamic in-vehicle environments.

[0030] In some embodiments of the present invention, S101 can be implemented through the following steps: the wireless channel state parameters include effective bandwidth and effective signal-to-noise ratio; the original coding bitrate corresponding to the next video slice to be played is obtained; the effective bandwidth, effective signal-to-noise ratio, and original coding bitrate are substituted into the formula for actual available transmission coding bitrate to obtain the actual available transmission coding bitrate; the first QoE influence factor is adaptively adjusted based on the actual available transmission coding bitrate to obtain the second QoE influence factor; the formula for actual available transmission coding bitrate is as follows: ; In the above formula, The actual usable transmission coding rate, The original encoded bitrate, For efficiency coefficient, For effective bandwidth, For effective signal-to-noise ratio.

[0031] Specifically, the actual available transmission bitrate is the maximum transmission capability that the next video slice can achieve under the wireless channel parameters of the currently playing video slice. The original encoded bitrate is the inherent data transmission rate corresponding to the next video slice to be played, which can be obtained by looking up a table in the precoding version library using the video content ID and slice number of the next video slice to be played. After mapping the physical layer's wireless channel parameters to the actual available transmission bitrate perceptible to the application layer using the actual available transmission bitrate formula, the first QoE influence factor is adjusted based on the actual available transmission bitrate.

[0032] In some embodiments of the present invention, the second QoE influence factor is obtained by adaptively adjusting the first QoE influence factor based on the actual available transmission coding bitrate through the following steps: the first QoE influence factor includes the original resolution and the original frame rate; the wireless channel parameters also include effective delay spread; the actual available transmission coding bitrate, the original resolution, and the original coding bitrate are substituted into the resolution adjustment formula to obtain the target resolution; the product between the preset frame rate protection threshold and the original coding bitrate is calculated; if the actual available transmission coding bitrate is not less than the product, the original frame rate is used as the target frame rate; if the actual available transmission coding bitrate is less than the product, the original frame rate, the preset frame rate protection threshold, and the actual available transmission coding bitrate are substituted into the frame rate adjustment formula to calculate the target frame rate; The resolution adjustment formula is as follows: ; In the above formula, For the target resolution, Original resolution; The frame rate adjustment formula is as follows: ; In the above formula, For the target frame rate, The original frame rate, This is the preset frame rate protection threshold.

[0033] Among them, the preset frame rate protection threshold can be .

[0034] In some embodiments of the present invention, the first QoE impact factor further includes the original total stutter time and the original initial buffer duration. Correspondingly, the second QoE impact factor is obtained by adaptively adjusting the first QoE impact factor based on the actual available transmission coding bitrate, which can be achieved through the following steps: obtaining the playback duration of the next video segment to be played; substituting the effective latency spread, playback duration, and original total stutter time into the total stutter time adjustment formula to calculate the target total stutter time; obtaining the cumulative data amount and latency overhead of the next video segment before playback; and substituting the cumulative data amount, latency overhead, and actual available transmission coding bitrate into the initial buffer duration adjustment formula to calculate the target initial buffer duration. The formula for adjusting the total lag time is as follows: ; In the above formula, The target is the total lag time. This represents the original total lag time. This is an empirical coefficient. For playback duration, , This is an empirical coefficient. The root mean square of the effective delay spread; The formula for adjusting the initial buffer duration is as follows: ; In the above formula, The target initial buffer duration, L For latency overhead, b This represents the cumulative amount of data.

[0035] Specifically, the playback duration can be found in the local pre-stored metadata database by using the video content ID of the next video slice to be played. L represents the latency overhead during the connection establishment process, which is an inherent attribute value that can be obtained by the manufacturer. The buffer duration of the next video slice to be played before playback is obtained by looking up the table. Then, the buffer duration and the original encoding bitrate are multiplied to calculate the data accumulation.

[0036] In some embodiments of the present invention, the calculation of the target sensitivity coefficients between multiple second QoE scores and first QoE scores in S104 can be achieved by the following steps: calculating the differences between multiple second QoE scores and first QoE scores to obtain individual differences; calculating the ratio between the individual differences and a preset perturbation amount to obtain the initial sensitivity coefficient corresponding to each first QoE influence factor; and using the absolute value of the initial sensitivity coefficient as the target sensitivity coefficient.

[0037] Specifically, the difference between each second QoE score and the first QoE score is calculated to obtain the individual difference corresponding to each second QoE score. Then, the ratio of each individual difference to a preset perturbation amount is calculated to obtain the initial sensitivity coefficient corresponding to each first QoE influence factor. The absolute value of the initial sensitivity coefficient is taken to finally obtain multiple target sensitivity coefficients.

[0038] The formula for calculating the target sensitivity coefficient is as follows: = ; In the above formula, For the first k The target sensitivity coefficient corresponding to the first QoE impact factor The parameter set for the target QoE scoring model. These are the normalized wireless channel parameters. The normalized eigenvector formed by the first QoE impact factor. For the first k The components corresponding to the second QoE impact factor k For indexing, First QoE rating, Preset disturbance amount.

[0039] Specifically, by quantifying the local partial derivatives of the output of the target QoE scoring model with respect to each normalized second QoE influence factor, the relative influence of different factors on user experience under the current wireless channel parameters is characterized, and the sensitivity coefficients corresponding to each second QoE influence factor are obtained.

[0040] In some embodiments of the present invention, the weight allocation based on the target sensitivity coefficient in S104 to obtain the user's preference prediction result for the first QoE influence factor can be achieved through the following steps: summing the target sensitivity coefficients to obtain the total sensitivity coefficient; substituting the target sensitivity coefficient and the total sensitivity coefficient into the normalization formula to calculate the user's predicted preference weight for each first QoE influence factor; and summing the predicted preference weights to obtain the preference prediction result. The normalization formula is as follows: ; In the above formula, For the first k The predicted preference weights corresponding to the first QoE impact factor k Index of the first QoE impact factor For the first k The target sensitivity coefficient corresponding to the first QoE impact factor The overall sensitivity coefficient is... K This represents the number of first QoE impact factors.

[0041] The preference prediction results can include each first QoE impact factor and its corresponding preference weight. The predicted preference weight is the order of importance that users attach to multiple first QoE impact factors when watching the next video segment to be played. The higher the preference weight, the more the user cares about the impact factor. Subsequently, the corresponding playback strategy can be set for the next video segment to be played based on the preference prediction results.

[0042] For example, preference prediction results can be derived from preference vectors. Each dimension of this vector is non-negative, and there are no restrictions on this.

[0043] In some embodiments of the present invention, the training process of the target QoE scoring model can be implemented through the following steps: under a preset ideal network environment, a first sample dataset is collected; the first sample dataset includes a fourth QoE influence factor and a first QoE scoring label corresponding to multiple independent video samples; the first QoE scoring model is trained using the first sample dataset to obtain the trained first QoE scoring model. In a dynamic vehicle environment, a second sample dataset is collected; the second sample dataset includes wireless channel parameter samples of the currently playing video slice sample and the fifth QoE influence factor of the next video slice sample to be played. The trained first QoE scoring model is used to predict the second sample dataset to obtain the second QoE scoring label; the second QoE scoring label is corrected using the wireless channel parameter samples to obtain the corrected third QoE scoring label; the corrected third QoE scoring label, the wireless channel parameter samples, and the fifth QoE influence factor are used to train the second QoE scoring model to obtain the trained target QoE scoring model.

[0044] Specifically, the ideal network environment is assumed to be one with sufficient bandwidth, stable signal, and low latency. In this scenario, a multi-user subjective experiment is conducted, which involves setting multiple independent video samples, each with a corresponding QoE influence factor. After watching the independent video samples, multiple users subjectively give a rating, which is used as the first QoE rating label. Then, the first QoE rating model is trained using the first sample dataset until the training conditions are met, resulting in the trained first QoE rating model.

[0045] In dynamic vehicular environments, such as typical vehicular propagation scenarios like densely populated urban areas, highways, and suburban roads, a second sample dataset is collected, including wireless channel parameter samples from a first video slice and the fifth QoE influence factor from a second video slice. The first and second video slices are adjacent and have a playback time sequence relationship. This second sample dataset is input into a pre-trained first QoE scoring model to obtain multiple predicted second QoE score labels. These labels are then corrected using wireless channel parameter samples (i.e., environmental state parameters) to obtain corrected third QoE score labels. The corrected third QoE score labels, wireless channel parameter samples, and the fifth QoE influence factor are used as an augmentation dataset to train the second QoE scoring model, resulting in a trained target QoE scoring model.

[0046] Specifically, when training the first QoE scoring model, the fourth QoE influence factor is concatenated to construct a four-dimensional feature vector. .in, Indicates resolution, Indicates frame rate, Indicates the total duration of buffering. The initial buffer duration is represented by the value. The four-dimensional feature vector can be preprocessed using the min-max normalization method. The preprocessed feature vector is then used to train the first QoE scoring model. This first QoE scoring model is a static QoE scoring model, which can be a neural network, such as a multilayer perceptron; there are no restrictions here.

[0047] In training the second QoE scoring model, the processing of the fifth QoE impact factor before input is the same as that of the fourth QoE impact factor. Compared to the first QoE scoring model, the second QoE scoring model is a dynamic, environment-adaptive QoE scoring model that can learn the correspondence between the QoE impact factor, wireless channel parameters, and QoE scores. The second QoE scoring model can be a neural network, such as a multilayer perceptron, long short-term memory network, or support vector machine, etc., without limitation.

[0048] The dynamic vehicle environment can be either the actual environment or a model environment.

[0049] In some embodiments of the present invention, the correction of the second QoE scoring label using wireless channel parameter samples to obtain the corrected third QoE scoring label can be achieved through the following steps: the wireless channel parameter samples include effective signal-to-noise ratio samples and effective delay spread samples; the reference signal-to-noise ratio, reference delay spread, effective signal-to-noise ratio samples, and effective delay spread samples are substituted into the QoE scoring correction formula to obtain the QoE scoring correction value; and the QoE scoring correction value and the second QoE scoring label are summed to obtain the third QoE scoring label. The QoE score correction formula is as follows: ; In the above formula, This is the corrected value for the QoE score. and Divided into empirical weighting coefficients, For reference signal-to-noise ratio, For reference delay spread, For effective delay expansion samples, This is a sample with an effective signal-to-noise ratio.

[0050] The QoE scoring correction formula characterizes the additional QoE offset caused by changes in signal-to-noise ratio and delay spread. The reference signal-to-noise ratio and reference delay spread can be the baseline values ​​obtained from multiple measurements in an ideal laboratory network environment.

[0051] Figure 2 This is a schematic diagram illustrating the effect of predicting the preferences of vehicle user 1 for different QoE influencing factors in a dynamic in-vehicle environment, as provided in an embodiment of the present invention. Figure 2 The system contains four identifiers, representing four first QoE influencing factors that affect the next video slice to be played: resolution, frame rate, total stuttering time, and initial buffering time. The horizontal axis represents normalized channel quality, i.e., the quality of wireless channel parameters, and the vertical axis represents normalized preference weight, i.e., the degree of importance that users attach to each first QoE influencing factor. Figure 3This is a schematic diagram illustrating the effect of a vehicle user 2's preference prediction results for different QoE influencing factors in a dynamic in-vehicle environment, as provided in an embodiment of the present invention.

[0052] Specifically, from Figure 2 and Figure 3 As can be seen, this invention can effectively characterize the dynamic correction process of user preference weights in in-vehicle video streaming scenarios as environmental conditions change. When the normalized channel quality is low, the normalized preference weights corresponding to smoothness-related factors such as total stuttering time and initial buffering time are relatively high. As the normalized channel quality gradually improves, the normalized preference weight corresponding to resolution continues to rise, while the normalized preference weight corresponding to initial buffering time gradually decreases, and the overall change in the preference weight corresponding to frame rate is relatively small.

[0053] Furthermore, comparing the output results of different user types reveals that User 1 and User 2 exhibit different characteristics during preference evolution. User 1's normalized resolution preference weight increases more rapidly and surpasses the normalized preference weight corresponding to total stuttering time in an earlier stage, indicating that this type of user shows a more significant increase in attention to image quality factors after environmental improvement. In contrast, User 2 maintains a higher normalized preference weight for total stuttering time across a wider range of normalized channel quality intervals, while the increase in the normalized preference weight for resolution is relatively gradual, indicating significant differences in the shift in emphasis between smoothness and image quality among different user types. Therefore, this invention not only outputs dynamically updated user preference prediction results that change with wireless channel parameters but also characterizes the differences in the magnitude, pace, and evolution of dominant factors among different user types, thereby achieving effective perception and characterization of user preferences in dynamic in-vehicle environments.

[0054] Figure 4 This is a schematic diagram of a video user preference prediction device in a dynamic vehicle environment, provided as an embodiment of the present invention. Figure 4 The device includes: Adjustment module 10 is used to adaptively adjust the first QoE impact factor of the next video slice to be played using the wireless channel parameters of the currently playing video slice, so as to obtain the second QoE impact factor. Prediction module 11 is used to predict the second QoE influence factor and wireless channel parameters using the target QoE scoring model to obtain the first QoE score of the next video slice to be played. The traversal and scoring module 12 is used to traverse the second QoE influence factor, add a preset perturbation amount to the current second QoE influence factor in each round to obtain the third QoE influence factor; use the target QoE scoring model to predict the third QoE influence factor, the remaining second QoE influence factors and the wireless channel parameters to obtain the second QoE score. After the traversal is completed, multiple second QoE scores are obtained. The calculation module 13 is used to calculate the target sensitivity coefficient between multiple second QoE scores and first QoE scores respectively, and to perform weight allocation based on the target sensitivity coefficient to obtain the user's preference prediction result for the first QoE influence factor.

[0055] This invention also provides an electronic device, such as... Figure 5 As shown, it includes a processor 001, a communication interface 002, a memory 003, and a communication bus 004. The processor 001, communication interface 002, and memory 003 communicate with each other via the communication bus 004. Memory 003 is used to store computer programs; When processor 001 executes the program stored in memory 003, it performs the following steps: The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0056] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0057] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0058] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0059] For devices / electronic devices, since they are basically similar to the method embodiments, the description is relatively simple, and relevant details can be found in the description of the method embodiments.

[0060] It should be noted that the electronic device in this embodiment of the invention is an electronic device that applies the above-described method for predicting user preferences for video streaming services in dynamic vehicle environments. Therefore, all embodiments of the above-described method for predicting user preferences for video streaming services in dynamic vehicle environments are applicable to this electronic device and can achieve the same or similar beneficial effects.

[0061] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0062] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0063] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus (devices), or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects, all of which are collectively referred to herein as "modules" or "systems." Furthermore, this application can take the form of a computer program product implemented 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. The computer program may be stored / distributed in a suitable medium, provided with or as part of other hardware, or may take other distribution forms, such as via the Internet or other wired or wireless telecommunications systems.

[0064] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products 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.

[0065] 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.

[0066] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus 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.

[0067] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for predicting video user preferences in a dynamic vehicle environment, characterized in that, include: The first QoE impact factor of the next video slice to be played is adaptively adjusted using the wireless channel parameters of the currently playing video slice to obtain the second QoE impact factor. The target QoE scoring model is used to predict the second QoE influence factor and wireless channel parameters to obtain the first QoE score of the next video slice to be played. The second QoE influence factor is iterated through, and a preset perturbation is superimposed on the current second QoE influence factor in each round to obtain the third QoE influence factor; the third QoE influence factor, the remaining second QoE influence factors, and the wireless channel parameters are predicted using the target QoE scoring model to obtain the second QoE score. After the iteration is completed, multiple second QoE scores are obtained. The target sensitivity coefficients between multiple second QoE scores and first QoE scores are calculated separately, and weights are assigned based on the target sensitivity coefficients to obtain the user's preference prediction results for the first QoE influence factor.

2. The method according to claim 1, characterized in that, The wireless channel state parameters include effective bandwidth and effective signal-to-noise ratio; The process of adaptively adjusting the first QoE impact factor of the next video slice to be played using the wireless channel parameters of the currently playing video slice to obtain the second QoE impact factor includes: Obtain the original bitrate of the next video slice to be played; Substituting the effective bandwidth, effective signal-to-noise ratio, and original coding rate into the formula for actual usable transmission coding rate, we obtain the actual usable transmission coding rate. The first QoE influence factor is adaptively adjusted based on the actual available transmission coding rate to obtain the second QoE influence factor; The formula for the actual usable transmission coding rate is as follows: ； In the above formula, The actual usable transmission coding rate, The original encoded bitrate, For efficiency coefficient, For effective bandwidth, For effective signal-to-noise ratio.

3. The method according to claim 2, characterized in that, The first QoE impact factor includes the original resolution and the original frame rate; the wireless channel parameters also include effective delay spread; The adaptive adjustment of the first QoE impact factor based on the actual available transmission coding rate to obtain the second QoE impact factor includes: The target resolution is obtained by substituting the actual available transmission coding bitrate, the original resolution, and the original coding bitrate into the resolution adjustment formula; Calculate the product between the preset frame rate protection threshold and the original encoding bitrate; if the actual available transmission encoding bitrate is not less than the product, then use the original frame rate as the target frame rate; if the actual available transmission encoding bitrate is less than the product, then substitute the original frame rate, the preset frame rate protection threshold, and the actual available transmission encoding bitrate into the frame rate adjustment formula to calculate the target frame rate. The resolution adjustment formula is as follows: ； In the above formula, For the target resolution, Original resolution; The frame rate adjustment formula is as follows: ； In the above formula, For the target frame rate, The original frame rate, This is the preset frame rate protection threshold.

4. The method according to claim 2, characterized in that, The first QoE impact factor includes the original total stutter time and the original initial buffer duration; The adaptive adjustment of the first QoE impact factor based on the actual available transmission coding rate to obtain the second QoE impact factor includes: Obtain the playback duration of the next video segment to be played; input the effective latency extension, playback duration, and original total stuttering time into the total stuttering time adjustment formula to calculate the target total stuttering time; Obtain the cumulative data amount and latency overhead of the next video slice to be played before playback; and substitute the cumulative data amount, latency overhead and actual available transmission coding bitrate into the initial buffer duration adjustment formula to calculate the target initial buffer duration; The formula for adjusting the total lag time is as follows: ； In the above formula, The target is the total lag time. This represents the original total lag time. This is an empirical coefficient. For playback duration, , This is an empirical coefficient. The root mean square of the effective delay spread; The formula for adjusting the initial buffer duration is as follows: ； In the above formula, The target initial buffer duration, L For latency overhead, b This represents the cumulative amount of data.

5. The method according to claim 1, characterized in that, The calculation of target sensitivity coefficients between multiple second QoE scores and first QoE scores includes: The individual differences are obtained by calculating the differences between multiple second QoE scores and first QoE scores; The ratios of individual differences and preset disturbances are calculated to obtain the initial sensitivity coefficients corresponding to each first QoE influence factor; the absolute value of the initial sensitivity coefficients is used as the target sensitivity coefficients.

6. The method according to claim 5, characterized in that, The weight allocation based on the target sensitivity coefficient to obtain the user's preference prediction result for the first QoE influence factor includes: The target sensitivity coefficients are summed to obtain the total sensitivity coefficient; the target sensitivity coefficient and the total sensitivity coefficient are substituted into the normalization formula to calculate the user's predicted preference weight for each first QoE influencing factor; and the predicted preference weights are summarized to obtain the preference prediction results. The normalization formula is as follows: ； In the above formula, For the first k The predicted preference weights corresponding to the first QoE impact factor k Index of the first QoE impact factor For the first k The target sensitivity coefficient corresponding to the first QoE impact factor The overall sensitivity coefficient is... K This represents the number of first QoE impact factors.

7. The method according to claim 1, characterized in that, The target QoE scoring model is trained through the following process: Under a pre-defined ideal network environment, a first sample dataset is collected; the first sample dataset includes the fourth QoE impact factor and the first QoE score label corresponding to multiple independent video samples; The first QoE scoring model is trained using the first sample dataset to obtain the trained first QoE scoring model. In a dynamic in-vehicle environment, a second sample dataset was collected. The second sample dataset includes wireless channel parameter samples of the first video slice sample and the fifth QoE influence factor of the second video slice sample; The first video slice sample and the second video slice sample are video slices that are temporally adjacent; The trained first QoE scoring model is used to predict the second sample dataset to obtain the second QoE scoring label; The second QoE score label is corrected using wireless channel parameter samples to obtain the corrected third QoE score label; The second QoE scoring model was trained using the corrected third QoE score label, wireless channel parameter samples, and the fifth QoE influence factor to obtain the trained target QoE scoring model.

8. The method according to claim 7, characterized in that, The wireless channel parameter samples include effective signal-to-noise ratio samples and effective delay spread samples; The step of correcting the second QoE scoring label using wireless channel parameter samples to obtain the corrected third QoE scoring label includes: Substitute the reference signal-to-noise ratio, reference delay spread, effective signal-to-noise ratio sample, and effective delay spread sample into the QoE score correction formula to obtain the QoE score correction value; then sum the QoE score correction value and the second QoE score label to obtain the third QoE score label. The QoE score correction formula is as follows: ； In the above formula, This is the corrected value for the QoE score. and Divided into empirical weighting coefficients, For reference signal-to-noise ratio, For reference delay spread, For effective delay expansion samples, This is a sample with an effective signal-to-noise ratio.

9. A video user preference prediction device for a dynamic vehicle-mounted environment, characterized in that, The device includes: The adjustment module is used to adaptively adjust the first QoE impact factor of the next video slice to be played by using the wireless channel parameters of the currently playing video slice, so as to obtain the second QoE impact factor. The prediction module is used to predict the second QoE influence factor and wireless channel parameters using the target QoE scoring model, so as to obtain the first QoE score of the next video slice to be played. The traversal and scoring module is used to traverse the second QoE influence factor, add a preset perturbation amount to the current second QoE influence factor in each round to obtain the third QoE influence factor; use the target QoE scoring model to predict the third QoE influence factor, the remaining second QoE influence factors and wireless channel parameters to obtain the second QoE score. After the traversal is completed, multiple second QoE scores are obtained. The calculation module is used to calculate the target sensitivity coefficients between multiple second QoE scores and first QoE scores, and to perform weight allocation based on the target sensitivity coefficients to obtain the user's preference prediction results for the first QoE influence factor.

10. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 8.

Images