High-definition wireless video stream data compression method based on adaptive code rate

Abstract

The application provides a high-definition wireless video stream data compression method based on adaptive code rate. In the scene where a wireless client cannot obtain a physical layer channel state and there is no server cooperation, only the time consumption and the buffer margin at the fragment boundary are used for selection. The method introduces a gear correction time consumption residual as a de-mixing discriminant, generates a tolerance threshold and a separation threshold by combining the buffer state, determines the interval landing point, and defines the residual fuzzy area between the two thresholds. The continuous fuzzy count and the fuzzy degree level are maintained in the interval, the lock level is updated when there is insufficient evidence, and the target code rate gear up is gated. The fragment boundary gating state structure is used to uniformly bear the discriminant, the threshold, the interval and the lock state information, so as to suppress the mis-downshift and the premature upshift, reduce the gear fluctuation and improve the gear selection stability.

Description

Technical Field

[0001] This invention relates to the field of audio and video transmission and adaptive bit rate control technology, and more specifically, to a method for high-definition wireless video stream data compression based on adaptive bit rate. Background Technology

[0002] Audio and video transmission and adaptive bitrate control technologies are widely used in segmented streaming media transmission scenarios under wireless networks. The wireless client side usually needs to determine the target bitrate level of the next segment based on network and buffer status at the segment boundary. Existing adaptive bitrate methods typically only use application layer feedback such as segment download time and buffer margin to estimate available throughput, and combine buffer thresholds and hysteresis rules to complete the bitrate control.

[0003] Existing technologies have shortcomings:

[0004] Existing technologies generally suffer from two main problems: First, download time is affected by both network fluctuations and changes in the target bitrate, making it difficult to distinguish the causes of download slowdowns and easily leading to erroneous downgrading or premature upgrading based on single observations. Second, the lack of continuous state accumulation and gating constraints for scenarios with insufficient evidence makes it difficult to stably suppress unnecessary jumps and maintain consistent bitrate selection. Therefore, under the aforementioned observational constraints, how to de-obfuscate the causes of download slowdowns at fragment boundaries and suppress misjudgments triggered by single observations has become a problem that urgently needs to be solved by those skilled in the art. To address these problems, this invention proposes a solution. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a high-definition wireless video stream data compression method based on adaptive bit rate to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The high-definition wireless video stream data compression method based on adaptive bitrate performs the following steps at each segment boundary:

[0008] S101: Obtain download feedback information and fragment boundary buffer status information; obtain the download time and target bitrate level of two adjacent fragments from the download feedback information, perform level correction on the fragment download time to remove the impact of fragment volume change caused by the change in target bitrate level, and obtain level correction time residual; determine the tolerance threshold by combining the fragment boundary buffer status information and the recent fluctuation intensity of the level correction time residual, and determine the separation threshold based on the tolerance threshold and the threshold spacing term that changes with the fragment boundary buffer status information; compare the level correction time residual with the tolerance threshold and the separation threshold to determine the interval landing point, which is used to indicate that the current fragment boundary is in one of the residual tolerance zone, the residual significantly deteriorated zone, or the residual ambiguous zone;

[0009] S102: Initialize the locking level to the unlocked state; when the interval landing point indicator is in the residual fuzzy area, maintain the continuous fuzzy count in the preset sliding window, and determine the fuzziness level based on the relative position of the gear correction time residual, the tolerance threshold, and the separation threshold; when the continuous fuzzy count and the fuzziness level are determined to be insufficient evidence, update the locking level to the locked state.

[0010] S103: Under the constraint of the locking level, determine the candidate target bitrate of the next segment based on the download feedback information and the segment boundary buffer status information and output it; wherein, when the locking level is locked, the target bitrate of the next segment is restricted to not be higher than the current target bitrate.

[0011] In a preferred embodiment, the download feedback information includes at least the download time and target bitrate tier for two adjacent segments; when the media duration of two adjacent segments is inconsistent, the download feedback information also includes the media duration of two adjacent segments; the tier correction time residual is obtained by performing target bitrate tier ratio correction on the download time ratio of two adjacent segments, and the media duration ratio is incorporated into the ratio correction when the media duration is inconsistent; the segment boundary buffer status information is used to provide buffer margin at the segment boundary.

[0012] In a preferred embodiment, the interval landing point is determined according to the following rules: when the residual of gear correction time is not greater than the tolerance threshold, the interval landing point is in the residual tolerance zone; when the residual of gear correction time is not less than the separation threshold, the interval landing point is in the residual significantly deteriorated zone; when the residual of gear correction time is greater than the tolerance threshold and less than the separation threshold, the interval landing point is in the residual ambiguity zone; and the separation threshold is higher than the tolerance threshold.

[0013] In a preferred embodiment, the recent volatility intensity is obtained according to the following rules: the residual sequence of gear correction time is aggregated on multiple consecutive segments covered by a preset sliding window, and the volatility metric is calculated based on the residual sequence; the preset sliding window length is an integer not less than 2; the volatility metric is at least one of standard deviation, root mean square deviation, and absolute deviation of mean.

[0014] In a preferred embodiment, the threshold spacing term is determined by the spacing reference value and the buffer adjustment coefficient. The buffer adjustment coefficient is obtained by comparing the buffer margin with the preset buffer safety threshold and the preset buffer risk threshold to obtain the buffer adjustment interval identifier, and is selected from multiple discrete buffer adjustment coefficient candidate values ​​according to the buffer adjustment interval identifier. The preset buffer safety threshold is higher than the preset buffer risk threshold, and the buffer adjustment coefficient selected when the buffer adjustment interval is a risk interval is not less than the buffer adjustment coefficient selected when the buffer adjustment interval is a safety interval.

[0015] In a preferred embodiment, the preset sliding window used to maintain the continuous fuzzy count is consistent with the sliding window used to calculate the intensity of recent fluctuations; the continuous fuzzy count is maintained according to the following rules: it is incremented when the interval landing point indicator is in the residual fuzzy region, and cleared to zero when the interval landing point indicator is in the residual tolerance region or the residual significantly deteriorated region.

[0016] In a preferred embodiment, the insufficient evidence is used to characterize a slight but persistent residual fuzziness condition and to trigger a lock level update to gate the target bit rate increment. Insufficient evidence is determined according to the following rules: when the continuous fuzziness count is greater than the number threshold and the fuzziness level is lower than the level threshold, it is determined to be insufficient evidence.

[0017] In a preferred embodiment, the level threshold is obtained by comparing the buffer margin with a preset buffer safety threshold and a preset buffer risk threshold to obtain a level threshold interval identifier, and then selecting from multiple discrete level threshold candidate values ​​according to the level threshold interval identifier. When the level threshold interval is a risk interval, the selected level threshold is not less than the level threshold selected when the level threshold interval is a safety interval. The number of times threshold is obtained by comparing the buffer margin with the preset buffer safety threshold and the preset buffer risk threshold to obtain a buffer state interval identifier, and then selecting from multiple discrete number of times candidate values ​​according to the buffer state interval identifier. When the buffer margin decreases, the selected number of times candidate value is not greater than the selected number of times candidate value when the buffer margin increases.

[0018] In a preferred embodiment, the ambiguity level is determined according to the following rules: the relative position is determined based on the ratio of the magnitude of the gear correction time residual exceeding the tolerance threshold to the distance between the tolerance threshold and the separation threshold, and the relative position is compared with multiple graded thresholds to determine multiple discrete levels, the multiple graded thresholds are arranged in ascending order of size; an increase in the relative position indicates that the gear correction time residual is closer to the separation threshold boundary within the residual ambiguity region.

[0019] In a preferred embodiment, the candidate next segment target bitrate tier is determined according to the following rules: the throughput capacity representation value is obtained based on the download feedback information, and a safety factor constraint is applied to the throughput capacity representation value in combination with the buffer margin; the highest tier in the target bitrate tier set that does not exceed the upper limit of the throughput capacity representation value after the safety factor constraint is selected as the candidate next segment target bitrate tier, wherein the safety factor is a positive number and less than 1, and the target bitrate tier set is a discrete tier list; when the locking level is locked and the candidate next segment target bitrate tier is higher than the current target bitrate tier, the current target bitrate tier is output.

[0020] The advantages and effects of the high-definition wireless video stream data compression method based on adaptive bitrate of this invention:

[0021] This invention introduces the residual time of bitrate correction as a de-obfuscation discriminant, and makes this discriminant, together with the fragment boundary buffer state, work on the generation of the tolerance threshold and the separation threshold. This creates an executable interval-based judgment entry point at the fragment boundary, ensuring a comparable judgment caliber even when download time is affected by both network fluctuations and changes in the target bitrate. Within the residual ambiguity region, this invention characterizes the evidence accumulation state through continuous ambiguity counting and ambiguity levels. When evidence is insufficient, it triggers a lock level update, using the lock level to gate the jump in the next fragment's target bitrate, ensuring that the next fragment's target bitrate in the locked state is not higher than the current target bitrate. Based on the above de-obfuscation discriminant caliber and gating constraint chain, it can suppress erroneous downgrading and premature upgrading triggered by a single observation, reduce repeated jumps and falls in the target bitrate under uncertain conditions, thereby reducing image quality fluctuations and improving the stability of fragment boundary bitrate selection. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0023] Figure 2 This is a schematic diagram of the observation extraction and residual interval determination process in step S101 of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention discloses a high-definition wireless video stream data compression method based on adaptive bitrate, applicable to scenarios where the wireless client cannot obtain the physical layer channel status and there is no server-side coordination. It performs adaptive bitrate selection control on the target bitrate level of the next segment at each segment boundary. In this scenario, the client can only observe the segment download time and buffer capacity, and the download time is simultaneously affected by network fluctuations and changes in the target bitrate level. This makes it difficult to distinguish the causes of slowdowns, easily leading to erroneous downgrading or premature upgrading triggered by a single observation.

[0026] The technical problem this invention aims to solve is how to deobfuscate the causes of slowdown downloads at fragment boundaries and suppress unstable rate jumps / ups / downs within the residual ambiguity region, under the aforementioned conditions of limited observation and causal confusion. To this end, this method introduces the residual time spent on rate correction as a deobfuscation criterion, and combines it with the fragment boundary buffer state to generate tolerance and separation thresholds for interval-based determination. The interval between the tolerance and separation thresholds where the residual time spent on rate correction lies is defined as the residual ambiguity region. Within the residual ambiguity region, this method triggers lock level updates through continuous ambiguity counting and ambiguity levels, and suppresses rate jumps in the next fragment's target bitrate level under lock level gating. Furthermore, this method uses the fragment boundary gating state structure G101 as a unified state carrier to consistently carry state information such as the residual time spent on rate correction, tolerance threshold, separation threshold, interval landing point, lock level, and the current target bitrate level.

[0027] Based on the above design, this invention constructs a complete process for a high-definition wireless video stream data compression method based on adaptive bit rate, consisting of steps S101 to S103 sequentially. (Refer to...) Figure 1 , Figure 1 This is a schematic diagram of the method flow of the present invention.

[0028] Step S101, De-obfuscation Discriminant Generation and Residual Interval Determination, is used to generate de-obfuscation discriminant quantities at the fragment boundaries and form interval determination entry points, completing the basic preparation for the discriminant generation stage of this method. This step reads the download feedback information X101, the fragment boundary buffer state information X102, and the fragment boundary selection control parameter set C101. First, it performs level correction on the fragment download time to obtain the level correction time residual. Then, it statistically analyzes the recent fluctuation intensity of the level correction time residual sequence within a sliding window, thereby generating the tolerance threshold and the separation threshold and determining the interval landing point. Finally, it writes the level correction time residual, tolerance threshold, separation threshold, and interval landing point into the fragment boundary gating state structure G101 for subsequent step S102 to read. Among them, X101 is used to provide information on fragment download time and target bitrate tier changes; X102 is used to provide buffer margin at fragment boundaries to participate in threshold generation; C101 is used to provide the sliding window and threshold generation settings; recent fluctuation intensity is used to characterize the fluctuation degree of the tier correction time residual sequence within the sliding window and participate in threshold generation; tier correction time residual is used to characterize the download time deviation after removing the influence of target bitrate tier changes; tolerance threshold is used to define the residual tolerance zone and residual ambiguity zone; separation threshold is used to define the residual ambiguity zone and residual significant degradation zone; interval landing point is used to indicate that the current fragment boundary is in one of the residual tolerance zone, residual ambiguity zone, or residual significant degradation zone; G101 is used to carry the de-obfuscation discrimination group and threshold and interval entry group to form a consistent downstream reading caliber.

[0029] Step S102 is used to maintain continuity evidence only in the residual ambiguity region at the fragment boundary and trigger a lock level update, forming a gating trigger chain that differs from conventional hysteresis patches. This step takes the fragment boundary buffer state information X102 and the fragment boundary selection control parameter set C101 as input, reads the interval landing point in the fragment boundary gating state structure G101, maintains the continuous ambiguity count and ambiguity level in the residual ambiguity region, dynamically determines the number threshold and level threshold, and completes the evidence insufficiency judgment. When the triggering condition is met, the lock level update result is written back to the fragment boundary gating state structure G101 for subsequent step S103 to read and gate the next fragment target bitrate level jump. Specifically, in this step, X102 is used to provide buffer margin to support the dynamic adjustment of the number threshold and the level threshold; C101 is used to provide a list of ambiguity level classification thresholds and adjustment rules for the number threshold and level threshold as the buffer changes; G101 serves as a unified state carrier to carry the interval landing point and write back the continuous ambiguity count, ambiguity level and lock level, and is used to maintain the preset sliding window for continuous ambiguity count, which is consistent with the sliding window used in step S101 to calculate the recent fluctuation intensity.

[0030] Special note: The ambiguity level is used to characterize the interval position of the gear correction time residual within the residual ambiguity region. A smaller ambiguity level indicates that the gear correction time residual is closer to the tolerance threshold boundary, the residual deviation is smaller, and the evidence of network degradation is weaker. For this type of slight but persistent ambiguity, if the buffer margin is high, conventional gear selection control is more likely to prematurely upshift, thus amplifying subsequent misjudgments and gear fluctuations. Therefore, step S102 limits insufficient evidence to a combination of consecutive ambiguity counts meeting a certain number of conditions and the ambiguity level being lower than the level threshold. When evidence is insufficient, the locking level is updated to gate the target bitrate gear increase in subsequent step S103. Conversely, when the ambiguity level is larger, it indicates that the gear correction time residual is closer to the separation threshold boundary, the evidence of network degradation is stronger, and the interval landing point is more likely to enter the significantly degraded residual region and follow the downshift processing chain. Therefore, the insufficient evidence locking chain is not used as the primary trigger.

[0031] Step S103, the next segment target bitrate tier decision and gating output, is used to apply lock level constraints to the candidate tier selection results at the segment boundary and complete the final tier selection output, so as to prevent the next segment target bitrate tier from jumping when locking is triggered in the residual ambiguity area. This step generates a throughput capacity characterization value and applies a safety coefficient constraint to determine the candidate next segment target bitrate tier based on the download feedback information X101, the segment boundary buffer state information X102, the segment boundary tier selection control parameter set C101, and the segment boundary gating state structure G101. At the same time, it reads the lock level and completes the gating judgment according to the prohibition of jumping logic, and finally outputs the next segment target bitrate tier for subsequent segment requests and downloads. In this step, X101 is used to generate a throughput capacity representation value to characterize the available throughput capacity of the next fragment; X102 is used to provide buffer margin to adjust the strength of the security factor constraint; C101 is used to provide the security factor and its adjustment rules as the buffer state changes, and to provide a discrete list of target bitrate tiers; G101 is used to provide the locking level, and to provide the current target bitrate tier written by the download feedback information X101 at the fragment boundary, as an upper limit reference for prohibiting up-hopping in the locked state.

[0032] Through the above process, this invention effectively suppresses erroneous downgrading and premature upgrading triggered by a single observation, under the condition that only the fragment download time and buffer margin can be observed, and the download time is affected by both network fluctuations and changes in the target bitrate.

[0033] The implementation process and operational effects of the method of the present invention will be described in detail below with reference to specific embodiments. It should be understood that the embodiments are only used to illustrate the technical solution of the present invention, and not to limit it. The relevant steps, parameters and module divisions can be appropriately adjusted without changing the essence of the invention.

[0034] In an optional embodiment, step S101 is used to generate a de-obfuscation discriminant at the segment boundary and form an interval-based decision entry. The discriminant and threshold boundaries are mapped to interval landing points and written into the segment boundary gating state structure G101, thereby providing a unified reading caliber for the residual fuzzy area accumulation and lock level update decision in the subsequent step S102. This step only generates and writes back the de-obfuscation discriminant group and the threshold and interval entry group of G101, without updating the residual fuzzy area accumulation group and lock level item of G101, so as to avoid overlapping with the accumulation update and lock level update responsibilities of step S102.

[0035] This step involves the extraction and alignment of observed measurements, generation of gear-level correction residuals, generation of sliding window fluctuation intensity, generation of buffer adjustment thresholds, determination of interval landing points, and writing back the results. The generation of residuals and recent fluctuation intensity includes gear-level correction residual generation and sliding window fluctuation intensity generation. The generation of thresholds and the determination of interval landing points and the writing back of results include buffer adjustment threshold generation and the determination of interval landing points and the writing back of results. (Refer to...) Figure 2 , Figure 2 This is a schematic diagram of the observation extraction and residual interval determination process of the present invention. The segment boundary gating state structure G101 serves as a unified state carrier at each segment boundary, used to carry the de-obfuscation discrimination quantity group, threshold and interval entry group, residual ambiguity region accumulation group, and lock state group. In this step, the first two groups are written to form a discrimination entry caliber that can be directly read downstream.

[0036] In the observation extraction and aperture alignment, the target bitrate level of the current segment is extracted from the download feedback information X101 as the current target bitrate level, and the current target bitrate level is written into the locked state group of the segment boundary gating state structure G101. When the media duration of two adjacent segments is inconsistent, the media duration of the two adjacent segments is extracted synchronously and used for equivalent volume aperture alignment, so that subsequent level correction considers the combined influence of the target bitrate level and the media duration on the segment volume. The buffer margin at the segment boundary is extracted from the segment boundary buffer state information X102. The window and measurement settings, including the sliding window length and fluctuation measurement type, are read from the segment boundary level selection control parameter set C101. The threshold generation settings, including the threshold coefficient, minimum tolerance threshold lower limit and threshold spacing settings, are read. The buffer boundary settings, including the buffer safety threshold and buffer risk threshold, are read.

[0037] In the generation of residuals and recent fluctuation intensity, the download time of two adjacent segments is corrected based on the two adjacent target bitrate changes in X101 to obtain the residual of the bitrate correction time. This residual is designed to isolate the influence of segment volume change caused by the target bitrate change as much as possible and can be directly compared with the threshold boundary. Then, the residual sequence of the bitrate correction time is aggregated on multiple consecutive segments according to the sliding window length N, and the recent fluctuation intensity is calculated according to the fluctuation metric type. The fluctuation metric type is kept consistent within the same playback session.

[0038] In the threshold and interval landing point determination and result write-back process, a tolerance threshold is generated based on the recent fluctuation intensity and threshold coefficient, and the tolerance threshold is not less than the minimum tolerance threshold lower limit. A separation threshold is obtained by superimposing a threshold spacing term on the tolerance threshold, wherein the threshold spacing term is determined by the spacing benchmark value and the buffer adjustment coefficient. The buffer adjustment coefficient is determined by the relative position of the buffer margin with respect to the buffer safety threshold and the buffer risk threshold, and the threshold spacing term increases when the buffer margin decreases. The gear correction time residual is compared with the tolerance threshold and the separation threshold to determine the interval landing point, so that the interval landing point indicates that the current segment boundary is in one of the residual tolerance zone, the residual ambiguity zone, or the residual significant deterioration zone. Finally, the gear correction time residual is written into the de-obfuscation discrimination group of G101, and the tolerance threshold, the separation threshold, and the interval landing point are written into the threshold and interval entry group of G101, so that step S102 can directly perform residual ambiguity zone accumulation and lock level update determination based on the write-back information in G101.

[0039] When the playback session is in the initial stage, making it impossible to obtain the download time and target bitrate tier information for two adjacent segments, or when the number of residual samples for tier correction time available for recent statistics is insufficient to support sliding window statistics, this step's output is generated using initialization and degradation calculation methods. For example, the tier correction time residual is set to 0, and the recent fluctuation intensity is set to 0 or calculated based on the obtained residual samples. The tolerance threshold is set to a default value not less than the lower limit of the minimum tolerance threshold. The separation threshold is obtained by combining the threshold spacing term with the tolerance threshold. The interval landing point is indicated by default as being within the residual tolerance zone, and the above default results are written into the threshold and interval entry group of G101. After the cumulative download feedback information meets the requirements for statistics for two adjacent segments and sliding window statistics, the tier correction time residual, recent fluctuation intensity, tolerance threshold, separation threshold, and interval landing point are generated again according to the aforementioned methods.

[0040] To facilitate implementation and standardize the generation of discrimination parameters, this embodiment provides an optional calculation scheme, and each symbol is given its meaning and engineering definition when it first appears.

[0041] set up Indicates the current fragment number; let... Indicates the first The download time for each segment is determined by the download feedback information. Given; suppose Indicates the first The target bitrate value corresponding to the target bitrate level of each segment is determined by... Given: When the durations of two adjacent media segments are inconsistent, assume... Indicates the first The media duration of each segment is determined by... Given; suppose This indicates the buffer clearance at the fragment boundary, determined by the fragment boundary buffer state information. Provided.

[0042] The download time residual after tier correction is used to characterize the deviation in download time after the impact of target bitrate tier changes has been largely eliminated. For example, it can be calculated as follows: First, define the tier correction ratio. When two adjacent media segments have the same duration, take... When media durations are inconsistent, the media duration ratio is incorporated into the ratio correction, and the result is taken as... Redefine gear position correction time residual ,in It is a non-negative scalar; the larger the value, the more obvious the deviation after correction and the more likely it is to trigger subsequent interval determination.

[0043] Let the sliding window length be... An integer not less than 2, used to limit the number of boundaries for recent statistical coverage. For example, it can be 8 to 20 to strike a tradeoff between suppressing occasional jitter and following network changes. This value and its update strategy are determined by the set of partition boundary selection control parameters. The default setting is to maintain consistency within the same playback session. Let the residual sequence covered by the window be... Let the volatility measurement function be... Its output is a non-negative scalar, which can be, for example, one of standard deviation, root mean square deviation, or absolute deviation of the mean, and the volatility metric type remains consistent within the same playback session. The recent volatility intensity can be, for example, defined as... .

[0044] The tolerance threshold is used to define the residual tolerance region and the residual ambiguity region. Let the threshold coefficient be... and Let the lower limit of the minimum tolerance threshold be... and Both are from Preset. An example of a definable tolerance threshold. .in To avoid oversensitivity caused by an excessively narrow threshold when the network is extremely stable or the sample size is too small, a threshold of 0.03 to 0.10 can be used as an example. This is used to adjust the amplification of the threshold with respect to the noise level. For example, values ​​from 1 to 3 can be used, with larger values ​​being more sensitive to fluctuations.

[0045] The separation threshold is used to define the region of residual ambiguity and the region of significant residual degradation. An example definition is provided. ,in This is the threshold spacing term, determined jointly by the spacing reference value and the buffer adjustment coefficient. Let the spacing reference value be... and ,Depend on By default, and for example, this value can be between 0.05 and 0.30, used to limit the upper scale of the blurred region width. Increasing the value results in a wider blurred region, thus reducing the probability of misjudgment in a single observation. Let the buffer safety threshold be... The buffer risk threshold is And satisfy Both are from Preset, example available Seconds The seconds represent the boundaries of sufficient buffering and near-stuttering, respectively. Let the cutoff function be... Indicates will Limiting the range to 0 to 1, first calculate the buffer margin ratio. To satisfy the monotonicity constraint that the threshold spacing term increases as the buffer margin decreases, a buffer adjustment coefficient can be taken as an example. and order Alternatively, to enhance the speed of weight boosting during a crisis, a nonlinear compression coefficient could be introduced, for example. and , Depend on By default, the example can be 2 to 6, and let Take again .

[0046] Interval point is used to map the residual to one of three intervals. Let the interval point be... For example, the mutual exclusion rule is determined as follows. :when hour, The indication is within the residual tolerance zone; when hour, The indicator is in the region of significant residual degradation; when hour, The indicator is in the residual ambiguity region.

[0047] Based on the above criteria, the steps Can Write to the fragment boundary gate state structure The decongestion discriminant group, and , and Write the threshold and interval entry group into G101 for subsequent steps. Directly read and execute the cumulative residual fuzzy area and lock level update determination.

[0048] In an optional embodiment, step S102 is used to maintain the continuous state quantity in the residual fuzzy region as a determineable number of times condition and evidence condition based on the interval determination entry provided by G101, and trigger the lock level update when the evidence is insufficient, so as to provide a directly readable gating basis for the subsequent step S103.

[0049] This step involves three actions and a write-back of results: entry point identification and state loading, residual fuzzy zone accumulation maintenance and fuzziness level determination, dynamic threshold determination and trigger judgment, and writing the accumulated results and lock level back to G101. First, the interval landing point in G101 is read to identify whether the current segment boundary is within the residual fuzzy zone. The continuous fuzz count and fuzziness level of the residual fuzzy zone accumulation group in G101 are loaded as the basis for this update, while the lock level is set to unlocked by default. When the interval landing point indicates that it is within the residual fuzzy zone, the continuous fuzz count is incrementally updated within a preset sliding window, and the fuzziness level is determined based on the relative position of the residual with time-consuming correction in G101 relative to the tolerance threshold and separation threshold. When the interval landing point indicates that it is within the residual tolerance zone or the residual significantly deteriorated zone, the continuous fuzz count is cleared to zero, and the lock level remains unlocked.

[0050] When the playback session is in the initial stage, resulting in the absence of a usable continuous fuzzy count and fuzziness level in the segment boundary gating state structure, or when the number of continuous samples of the currently accumulated residual fuzzy region is insufficient to cover the preset sliding window length, the continuous fuzzy count is set to zero and the locking level is kept in the unlocked state. At the same time, the count condition is determined to be invalid and the insufficient evidence condition is determined to be invalid. This initialization state is written back to the residual fuzzy region accumulation group and the locking state group of the segment boundary gating state structure. After the preset sliding window length is met, the continuous fuzzy count and fuzziness level are updated according to the aforementioned criteria, and the insufficient evidence trigger judgment is executed.

[0051] Subsequently, based on the buffer margin in X102 and the adjustment rules in C101, the count threshold and grade threshold are determined. The count threshold decreases when the buffer margin decreases, and the grade threshold increases when the buffer margin decreases. The count condition is determined by comparing the continuous fuzzy count with the count threshold, and the evidence is determined by comparing the fuzziness grade with the grade threshold. If the continuous fuzzy count is greater than the count threshold and the fuzziness grade is lower than the grade threshold, the evidence is deemed insufficient. When the interval landing point indication is in the residual fuzzy region, the count condition is met, and the evidence is insufficient, the locking level is updated to the locked state and written back to the locked state group in G101. Simultaneously, the continuous fuzzy count and fuzziness grade are written back to the residual fuzzy region accumulation group in G101, allowing step S103 to directly read the locking level. In the locked state, the target bitrate of the next segment output is ensured to be no higher than the current target bitrate.

[0052] In an optional implementation, the relative position is defined. This represents the normalized offset of the gear shift correction time residual between the tolerance threshold and the separation threshold, where... This is a non-negative real number; a larger value indicates that the value is closer to the separation threshold boundary. The relative position is calculated as follows: ;in For the first The time consumption of residual correction at the boundary of each segment. For the first The tolerance threshold for each partition boundary For the first The separation threshold of each segment boundary, and When the calculation yields Set it to 0 when it is less than 0, and when When it is greater than 1, set it to 1, so that It falls within the range of 0 to 1.

[0053] Furthermore, a list of hierarchical thresholds arranged in ascending order of size is read from the segmentation boundary selection control parameter set C101. ,in Indicates the first Each tiered threshold This indicates the number of grading thresholds. It defines the discrete fuzziness levels. The discrete index used to characterize the degree of ambiguity. Take a positive integer, and The larger the value, the higher the ambiguity. Compare with the list of tiered thresholds to determine The determination rule is: when Time to take ;when Time to take ;when Time to take For ease of implementation, the tiered threshold list can use example values ​​divided into equally spaced segments. For instance, when setting four ambiguity levels, the following values ​​could be used: and order , , The reason for using equally spaced segments is that it is simple to implement and satisfies monotonicity, allowing the ambiguity level to change with... A monotonically increasing ambiguity level facilitates subsequent comparison with the level threshold to trigger a determination of insufficient evidence. Specifically, an increase in the ambiguity level indicates that the residual of the gear correction time is closer to the separation threshold boundary within the residual ambiguity region, which is used to characterize the interval where network-side degradation evidence is stronger.

[0054] In an optional embodiment, step S103 converts the locking level in G101 into an up-jump constraint on the candidate next segment target bitrate, and outputs the next segment target bitrate gated by the constraint at the segment boundary for direct use by subsequent segment requests and downloads.

[0055] This step involves generating a throughput capacity representation value and constraining a safety factor, determining the target bitrate tier for the candidate next segment, and determining the lock level gating and final output. First, this step generates a throughput capacity representation value from the download feedback information X101, which is used to estimate the available throughput capacity of the next segment. This throughput capacity representation value can be calculated from the download time of the most recent segment and the corresponding segment size representation, and the updated value remains consistent within the same playback session. Next, the buffer margin is obtained from the segment boundary buffer state information X102, and a safety factor constraint is applied to the throughput capacity representation value based on the safety factor in the segment boundary tier selection control parameter set C101 and its adjustment rules as the buffer state changes, to obtain the upper limit of the throughput capacity representation value after the safety factor constraint. In an optional implementation, the safety factor value decreases when the buffer margin decreases, making the target bitrate tier for the candidate next segment more conservative. Subsequently, this step selects the highest discrete bitrate in the target bitrate set that does not exceed the upper limit of the throughput capacity representation value after the security coefficient constraint as the candidate next segment target bitrate ...

[0056] In an optional implementation, a throughput capacity representation value is defined. For the first Each partition boundary can be estimated using throughput capacity. Calculated based on the downloaded feedback information X101 according to the preset specifications. Define the safety factor. is a conservative coefficient applied to the throughput capacity characterization value, where It is a positive number and less than 1. A smaller value indicates a more conservative approach. The set of control parameters for segmentation boundary selection, C101, includes a list of candidate values ​​for the preset discrete safety factor. ,in Indicates the first One candidate value for the safety factor. Indicates the number of candidate values. Based on buffer margin. With buffer safety threshold and buffer risk threshold The comparison yields the buffer state interval identifier, and the safety factor is selected from the list of discrete safety factor candidate values ​​based on the buffer state interval identifier. And the safety factor selected when the buffer margin decreases. The value should not exceed the safety factor selected when the buffer margin increases. Define the upper limit of the throughput capacity characterization value after the safety factor constraint. The calculation method for the upper limit of the target bitrate for selecting the next candidate segment is as follows: And select no more than from the target bitrate set. The highest discrete bit rate is used as the candidate target bit rate bit rate for the next segment.

[0057] For ease of implementation, the list of candidate values ​​for the discrete safety factor can use example values. For example, the following can be taken: and order , , The reason for setting discrete candidate values ​​is to avoid frequent changes in the safety factor due to small fluctuations in the buffer, thereby reducing gear selection jitter and enabling a more conservative upper limit of candidate gears when the buffer is low by selecting a smaller safety factor.

[0058] This step outputs the target bitrate of the next segment as the result of this step, and uses this target bitrate of the next segment as the target bitrate during subsequent segment requests and downloads to perform adaptive bitrate selection control, and writes the current target bitrate given by the download feedback information X101 into the locked state group of G101.

[0059] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0060] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0061] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0063] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A high-definition wireless video stream data compression method based on adaptive bit rate, characterized in that, Perform the following steps at each fragment boundary: S101: Obtain download feedback information and fragment boundary buffer status information; obtain the download time and target bitrate level of two adjacent fragments from the download feedback information, perform level correction on the fragment download time to remove the impact of fragment volume change caused by the change in target bitrate level, and obtain level correction time residual; determine the tolerance threshold by combining the fragment boundary buffer status information and the recent fluctuation intensity of the level correction time residual, and determine the separation threshold based on the tolerance threshold and the threshold spacing term that changes with the fragment boundary buffer status information; compare the level correction time residual with the tolerance threshold and the separation threshold to determine the interval landing point, which is used to indicate that the current fragment boundary is in one of the residual tolerance zone, the residual significantly deteriorated zone, or the residual ambiguous zone; S102: Initialize the locking level to the unlocked state; when the interval landing point indicator is in the residual fuzzy area, maintain the continuous fuzzy count in the preset sliding window, and determine the fuzziness level based on the relative position of the gear correction time residual, the tolerance threshold, and the separation threshold; when the continuous fuzzy count and the fuzziness level are determined to be insufficient evidence, update the locking level to the locked state. S103: Under the constraint of the locking level, determine the candidate target bitrate of the next segment based on the download feedback information and the segment boundary buffer status information and output it; wherein, when the locking level is locked, the target bitrate of the next segment is restricted to not be higher than the current target bitrate.

2. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The download feedback information includes at least the download time and target bitrate tier for two adjacent segments; when the media duration of two adjacent segments is inconsistent, the download feedback information also includes the media duration of the two adjacent segments; the residual time for tier correction is obtained by performing target bitrate tier ratio correction on the ratio of download time of two adjacent segments, and the media duration ratio is incorporated into the ratio correction when the media duration is inconsistent; the segment boundary buffer status information is used to provide buffer margin at the segment boundary.

3. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The interval landing point is determined according to the following rules: when the residual of gear correction time is not greater than the tolerance threshold, the interval landing point is in the residual tolerance zone; when the residual of gear correction time is not less than the separation threshold, the interval landing point is in the residual significantly deteriorated zone; when the residual of gear correction time is greater than the tolerance threshold and less than the separation threshold, the interval landing point is in the residual ambiguity zone; and the separation threshold is higher than the tolerance threshold.

4. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The recent volatility intensity is obtained according to the following rules: the residual sequence of gear correction time is aggregated on multiple consecutive segments covered by a preset sliding window, and the volatility metric is calculated based on the residual sequence; the preset sliding window length is an integer not less than 2; the volatility metric is at least one of standard deviation, root mean square deviation, and absolute deviation of mean.

5. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The threshold spacing term is determined by the spacing benchmark value and the buffer adjustment coefficient. The buffer adjustment coefficient is obtained by comparing the buffer margin with the preset buffer safety threshold and the preset buffer risk threshold to obtain the buffer adjustment interval identifier. Based on the buffer adjustment interval identifier, it is selected from multiple discrete buffer adjustment coefficient candidate values. The preset buffer safety threshold is higher than the preset buffer risk threshold, and the buffer adjustment coefficient selected when the buffer adjustment interval is a risk interval is not less than the buffer adjustment coefficient selected when the buffer adjustment interval is a safety interval.

6. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The preset sliding window used to maintain continuous fuzzy counts is the same as the sliding window used to calculate recent fluctuation intensity; continuous fuzzy counts are maintained according to the following rules: the interval landing point indicator is incremented when it is in the residual fuzzy region, and cleared to zero when it is in the residual tolerance region or the residual significantly deteriorated region.

7. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The lack of evidence is used to characterize a slight but persistent residual fuzziness condition and to trigger a lock level update to gate the target bit rate increment. The lack of evidence is determined according to the following rules: when the continuous fuzziness count is greater than the number threshold and the fuzziness level is lower than the level threshold, it is determined to be insufficient evidence.

8. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 7, characterized in that, The level threshold is obtained by comparing the buffer margin with a preset buffer safety threshold and a preset buffer risk threshold to obtain a level threshold interval identifier, and then selecting from multiple discrete level threshold candidate values ​​according to the level threshold interval identifier. When the level threshold interval is a risk interval, the selected level threshold is not less than the level threshold selected when the level threshold interval is a safety interval. The number of times threshold is obtained by comparing the buffer margin with a preset buffer safety threshold and a preset buffer risk threshold to obtain a buffer state interval identifier, and then selecting from multiple discrete number of times candidate values ​​according to the buffer state interval identifier. When the buffer margin decreases, the selected number of times candidate value is not greater than the selected number of times candidate value when the buffer margin increases.

9. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The ambiguity level is determined according to the following rules: the relative position is determined based on the ratio of the magnitude of the residual error of the gear correction time exceeding the tolerance threshold to the distance between the tolerance threshold and the separation threshold, and the relative position is compared with multiple graded thresholds to determine multiple discrete levels, and the multiple graded thresholds are arranged in ascending order of size; An increase in relative position indicates that the residual time consumed by gear correction is closer to the separation threshold boundary within the residual ambiguity region.

10. The high-definition wireless video stream data compression method based on adaptive bit rate according to claim 1, characterized in that, The target bitrate of the candidate next segment is determined according to the following rules: the throughput capacity characterization value is obtained based on the download feedback information, and a safety factor constraint is applied to the throughput capacity characterization value in combination with the buffer margin; Select the highest bitrate from the target bitrate set that does not exceed the upper limit of the throughput capacity representation value after the safety factor constraint as the candidate target bitrate of the next segment, where the safety factor is a positive number and less than 1, and the target bitrate set is a discrete bitrate list; when the locking level is locked and the candidate target bitrate of the next segment is higher than the current target bitrate, output the current target bitrate.

Images