A stream media service fault-tolerant method and system supporting pre-play and post-push and discontinuous push

Abstract

The application belongs to the technical field of information technology, and specifically discloses a streaming media service fault-tolerant method and system supporting pre-broadcast post-push and discontinuous push, wherein the method comprises the following steps: acquiring current buffer data receiving speed and end-to-end delay accumulation value from streaming media transmission data by monitoring network fluctuation and high packet loss rate in real time, processing the values by using Kalman filtering algorithm to filter noise, and obtaining a smoothed network feature vector; judging whether a chain reaction occurs according to the obtained smoothed network feature vector, determining that there is a potential interruption risk if the delay accumulation exceeds a preset threshold and the buffer data receiving speed is lower than the average level, and then generating a risk assessment report; and the application solves the problem of potential interruption risk caused by network fluctuation and high packet loss rate in streaming media transmission in the prior art.

Description

Technical Field

[0001] This invention relates to the technical field of information technology, specifically to a fault-tolerant method and system for streaming media services that supports both pre-broadcast and post-continuous streaming. Background Technology

[0002] Streaming media services have become a core pillar of modern internet entertainment and real-time communication, and their stability and continuity directly determine the quality of the user experience. Given the complex and ever-changing network environment, ensuring uninterrupted transmission of video or audio content has become a key issue of continuous concern for the industry.

[0003] Many current streaming media systems are often optimized for a single transmission mode in their fault tolerance design, making it difficult to balance smooth playback and data reliability in real-world mixed scenarios. When network fluctuations occur, the system easily falls into a dilemma: either prematurely interrupt playback to wait for data recovery, resulting in excessively long perceived buffering times for users; or forcibly maintain playback, only to suffer severe black screens or audio interruptions due to buffer exhaustion. This single-response strategy lacks flexibility and specificity in its fault tolerance behavior when facing network jitter, switching, or brief interruptions, making it difficult to simultaneously meet the typical business requirements of both play-before-push and interrupted-continuous-push.

[0004] A deeper challenge lies in the difficulty of adapting fault tolerance trigger criteria to dynamically changing network and playback conditions. There is a close interplay between network packet loss rate, buffer fill speed, and end-to-end latency: an increased packet loss rate directly slows down the arrival speed of data in the buffer, thus increasing actual latency; and accumulated latency, in turn, forces the system to attempt retransmissions or reconnections more frequently, further exacerbating network load. This chain reaction makes it impossible for fixed thresholds to accurately distinguish between brief fluctuations and sustained degradation, nor can it reasonably balance sensitivity and stability across different service types. For example, in a play-before-push scenario, triggering the protection mechanism based solely on a single packet loss rate threshold might prematurely pause push notifications when network fluctuations are minimal, wasting accumulated buffer space; while in intermittent push scenarios, if the reconnection interval is set too long, multiple short interruptions will lead to a sharp decline in user experience without timely network switching or strategy adjustments.

[0005] Therefore, in streaming media services where pre-broadcast and post-push coexist, establishing an adaptive fault-tolerant triggering mechanism that can comprehensively consider network quality, buffer status, and latency characteristics has become a key issue in ensuring service continuity and user experience. Summary of the Invention

[0006] This invention provides a fault-tolerant method and system for streaming media services that supports both pre-broadcast and post-disconnection push, aiming to solve the problem of potential interruption risks caused by network fluctuations and high packet loss rates in existing streaming media transmission technologies.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A fault-tolerant method for streaming media services supporting both pre-broadcast and post-interruption continuous push modes includes: real-time monitoring of network fluctuations and high packet loss rates; obtaining the current buffered data reception speed and end-to-end latency accumulation from the streaming media transmission data; processing these values ​​using a Kalman filter algorithm to filter noise and obtain a smoothed network feature vector; determining whether a chain reaction has occurred based on the obtained smoothed network feature vector; if the latency accumulation exceeds a preset threshold and the buffered data reception speed is below average, identifying a potential interruption risk and generating a risk assessment report; and analyzing the transmission mode in a mixed scenario based on the generated risk assessment report, using a dynamic programming algorithm to optimize the switching path between pre-broadcast and post-interruption continuous push modes to obtain an optimized mode switching sequence. The process involves several steps: First, by obtaining the mode switching sequence, the activation condition of the adaptation mechanism is triggered. If the smoothing network feature vector shows continuous fluctuations, the buffer filling strategy is adjusted to maintain smooth playback. Based on the adjusted buffer filling strategy, feedback data from reliable transmission is obtained to determine whether lost packets need to be retransmitted. If the feedback data indicates a high packet loss rate, a retransmission operation is performed and the latency accumulation record is updated. The updated latency accumulation record is used to determine the degree of mitigation of the chain reaction. The smoothing network feature vector is recalculated using a Kalman filter algorithm to obtain the corrected index value. Based on the corrected index value, the transmission mode is optimized in mixed scenarios to ensure continuity. It is then determined whether smooth playback has been restored. If not, the adaptation mechanism is repeatedly triggered until reliable transmission stabilizes.

[0009] In one aspect of the invention, the step of obtaining the current buffered data reception speed and end-to-end latency accumulation value from streaming media transmission data by real-time monitoring of network fluctuations and high packet loss rates, and processing these values ​​using a Kalman filter algorithm to filter noise, to obtain a smoothed network feature vector, includes:

[0010] By monitoring network fluctuations and high packet loss rates in real time, the buffered data reception speed and latency accumulation value are extracted from the streaming media transmission data to obtain the initial network status data.

[0011] Based on the initial network state data, the Kalman filter algorithm is used to filter noise from the buffered data receiving speed and the accumulated delay value to obtain a smoothed data sequence.

[0012] For the smoothed data sequence, analyze the changing trend of the smoothing network feature vector to determine whether there are significant fluctuations or delay accumulation phenomena.

[0013] If the analysis results show that the smooth network feature vector has significant fluctuations or accumulated delays, the transmitted data is prioritized to obtain the processing requirements of key data streams and determine whether to adjust the transmission strategy.

[0014] By assessing the processing requirements of key data streams and combining the results of real-time monitoring, an optimized transmission path is determined, resulting in an adjusted transmission configuration.

[0015] Based on the adjusted transmission configuration, update the buffer data receiving speed and latency accumulation control parameters for streaming media transmission, obtain the latest smooth network feature vector data, and complete the optimization process.

[0016] In one aspect of the invention, the step of determining whether a chain reaction has occurred based on the obtained smooth network feature vector, and if the accumulated delay exceeds a preset threshold and the buffered data reception speed is lower than the average level, then a potential interruption risk is determined, and a risk assessment report is generated, including:

[0017] By monitoring the feature vectors of the smoothing network in real time, dynamic data on latency accumulation and buffered data reception speed are obtained. These data are then used for preliminary screening to identify key points of abnormal fluctuations.

[0018] Based on the key points of abnormal fluctuations, analyze the specific distribution of time delay accumulation. If the distribution exceeds the preset threshold, it is determined that there is a potential problem and a corresponding abnormality mark is generated.

[0019] For the generated anomaly markers, cross-validation is performed using preset comparison rules in conjunction with the deviation data of buffered data receiving speed to determine whether the speed deviation has a correlation with the accumulation of time delay.

[0020] Based on the analysis of the correlation between speed deviation and time delay accumulation, the triggering conditions for interruption risk are extracted. If the triggering conditions meet the preset standards, a risk warning signal is generated.

[0021] Based on the risk warning signal, the data analysis module is invoked to trace the root cause of potential problems, identify the main factors affecting network quality, and determine key intervention points;

[0022] For key intervention points, the priority configuration of transmission paths is adjusted, and optimized network quality data is obtained by accumulating status changes through real-time monitoring.

[0023] Based on the optimized network quality data, the information on risk assessment is updated, and by continuously tracking changes in indicators, it is determined whether there are any new potential problems.

[0024] In one aspect of the invention, the analysis of transmission modes in a mixed scenario based on the generated risk assessment report, and the optimization of the switching path between pre-broadcast and post-discontinuous push using a dynamic programming algorithm to obtain an optimized mode switching sequence, includes:

[0025] Obtain the accumulated time delay value and the buffered data reception speed value from the current smooth network feature vector data;

[0026] By comparing the accumulated delay value with a preset threshold, if the value exceeds the threshold and the buffered data reception speed is lower than the benchmark value, the interruption risk point is determined.

[0027] For points at risk of interruption, obtain the current transmission mode status and determine whether it is a case of broadcast first and then push or interruption followed by continuous push.

[0028] Obtain the most recent mode switching history sequence from the transmission mode status;

[0029] A dynamic programming algorithm is used to process the historical sequence of mode switching, calculate the total cost of each switching path, and obtain the optimized mode switching sequence corresponding to the path with the lowest cost.

[0030] Perform dynamic switching operations on the transmission mode according to the optimization mode switching sequence, and update the transmission mode status;

[0031] By dynamically switching the transmission mode status, new smooth network feature vector data is obtained to determine whether the interruption risk points have disappeared.

[0032] In one aspect of the invention, the step of triggering the activation condition of the adaptation mechanism through the obtained mode switching sequence, and adjusting the buffer filling strategy to maintain smooth playback if the smoothing network feature vector shows continuous fluctuations, includes:

[0033] By using the mode switching sequence, monitoring data on the current network quality is obtained. Based on the fluctuation of the indicators, it is determined whether the triggering conditions are met, and a preliminary judgment on the response requirements is obtained.

[0034] If the initial response to the demand indicates that there are fluctuations in network quality indicators, then the specific fluctuation frequency and amplitude are extracted from the monitoring data, and a preset threshold is used for comparison to determine whether to activate the adaptation mechanism.

[0035] Based on the activation status of the adaptation mechanism, the current buffer filling level is obtained. In the case of insufficient buffer filling, the filling rate adjustment parameter is calculated to obtain the direction of strategy adjustment.

[0036] By adjusting the direction of the strategy, we can dynamically respond to the actual needs of buffer filling, adjust the filling rate and priority allocation, and determine the data replenishment rhythm of the buffer.

[0037] Based on the data replenishment rhythm of the buffer, we continuously monitor changes in network quality and track the persistence of indicator fluctuations in real time to ensure the continuity of playback.

[0038] By playing the continuous guarantee status and combining the results of quality monitoring, we can dynamically respond to whether there is a decline in network quality and determine whether to maintain the current strategy adjustment plan.

[0039] Based on the maintenance status of the strategy adjustment plan, obtain the latest switching sequence data, record the frequency and effect of mode switching, and determine the basis for subsequent dynamic response optimization.

[0040] In one aspect of the invention, the step of obtaining reliable transmission feedback data according to the adjusted buffer filling strategy, determining whether lost packets need to be retransmitted, and performing a retransmission operation and updating the latency accumulation record if the feedback data indicates a high packet loss rate, includes:

[0041] Based on the adjusted buffer filling strategy, reliable transmission feedback data is obtained to determine whether the packet loss rate exceeds the preset threshold. If the packet loss rate is high, lost data packets are marked and retransmission requests are triggered to obtain retransmission queue records.

[0042] By recording the retransmission queue, the retransmission delay data of each lost packet is obtained, the cumulative delay increment is calculated, and the current playback delay status is determined.

[0043] Based on the current playback delay status, obtain the video frame decoding timestamp difference to determine whether decoding starvation has occurred. If decoding starvation has occurred, adjust the data packet decoding priority to obtain an emergency frame replacement sequence.

[0044] By using emergency frame interpolation sequences, the continuity of sequence numbers of received data packets is detected, the position of sequence number gaps is extracted, and the list of actual gap positions is determined.

[0045] Based on the list of actual gap locations, filter out the key frame packets that have not yet arrived from the reliable transmission feedback data, mark them as high-priority retransmission targets, and obtain the key retransmission subset;

[0046] By dynamically updating the retransmission window size and retransmission timer duration using key retransmission subsets, a new round of retransmission rhythm is determined.

[0047] Based on the new round of retransmission schedule, we continuously acquire the latest reliable transmission feedback data, update the cumulative latency increment and packet loss rate statistics in real time, and obtain the transmission quality assessment results.

[0048] In one aspect of the invention, the step of determining the degree of mitigation of the chain reaction through updated time delay accumulation records, and recalculating the smoothing network feature vector using a Kalman filter algorithm to obtain a corrected index value, includes:

[0049] By accumulating and recording latency data, latency variation data in network transmission is extracted, and the fluctuation of latency variation is classified to obtain a preliminary fluctuation distribution.

[0050] Based on the preliminary fluctuation distribution, the correlation between network transmission and data stability is analyzed. If the fluctuation distribution exceeds the preset threshold, the network status is marked to determine the abnormal fluctuation range.

[0051] By identifying abnormal fluctuation ranges, quality index data under network conditions is obtained. The Kalman filter algorithm is then used to smooth the quality indexes, resulting in a corrected evaluation result.

[0052] Based on the revised assessment results, the frequency of indicator updates is adjusted, and real-time monitoring of dynamic changes in transmission performance is conducted to determine the current stability level of network transmission.

[0053] By analyzing the degree of matching between delay changes and data stability through the stability level, if the degree of matching is lower than the preset threshold, the sampling interval of dynamic monitoring is adjusted to obtain the optimized monitoring rhythm.

[0054] Based on the optimized monitoring schedule, the latest network status data is continuously acquired, and compared and analyzed with the evaluation results to determine the long-term fluctuation trend of network transmission.

[0055] In one aspect of the invention, the step of optimizing the transmission mode in a mixed scenario to ensure continuity based on the corrected index value, determining whether playback smoothness has been restored, and if not, repeatedly triggering an adaptation mechanism until reliable transmission is stable includes:

[0056] Obtain the corrected index value and the current mixed scene parameters;

[0057] Based on the corrected index values ​​and mixed scenario parameters, determine the transmission mode switching strategy;

[0058] A path switching strategy is used to perform path switching, resulting in a delayed sequence after the switch.

[0059] The delayed sequence is processed by the Kalman filter algorithm to obtain a smoothing quality estimate.

[0060] The smoothness of playback is determined based on the smoothness quality estimate.

[0061] If the playback is not stable, the transmission mode switching strategy will be redefined.

[0062] Based on the smoothness quality estimate and playback smoothness status, the transmission path changes are continuously monitored until a reliable and stable state is achieved.

[0063] In another aspect, the present invention also relates to a fault-tolerant system for streaming media services that supports both pre-broadcast and post-continuous streaming, the system comprising:

[0064] The real-time monitoring and data acquisition module is used to obtain the current buffered data reception speed and latency accumulation value from streaming media transmission data by monitoring network fluctuations and high packet loss rates in real time.

[0065] The smoothing network feature vector processing module is used to process these values ​​using the Kalman filter algorithm to filter out noise and obtain smoothed network feature vectors.

[0066] The chain reaction judgment and risk assessment module is used to determine whether a chain reaction has occurred based on the obtained smooth network feature vector. If the accumulated delay exceeds the preset threshold and the buffered data reception speed is lower than the average level, it is determined that there is a potential interruption risk, and then a risk assessment report is generated.

[0067] The transmission mode analysis and optimization module is used to analyze the transmission mode in a mixed scenario based on the generated risk assessment report. It uses a dynamic programming algorithm to optimize the switching path between pre-broadcast and post-continuous push, and obtains an optimized mode switching sequence.

[0068] The adaptation mechanism triggering and buffer adjustment module is used to trigger the activation condition of the adaptation mechanism based on the obtained mode switching sequence. If the smoothing network feature vector shows continuous fluctuations, the buffer filling strategy is adjusted to maintain smooth playback.

[0069] The feedback data processing and retransmission module is used to obtain reliable transmission feedback data based on the adjusted buffer filling strategy, determine whether lost packets need to be retransmitted, and if the feedback data indicates a high packet loss rate, then a retransmission operation is performed and the latency accumulation record is updated.

[0070] The chain reaction mitigation module is used to determine the degree of mitigation of the chain reaction by updating the accumulated time delay records, and to recalculate the smoothing network feature vector using the Kalman filter algorithm to obtain the corrected index value.

[0071] The indicator correction module is used to optimize the transmission mode in mixed scenarios to ensure continuity based on the corrected indicator values, determine whether playback smoothness has been restored, and if not, repeatedly trigger the adaptation mechanism until reliable transmission is stable.

[0072] Compared with the prior art, the present invention has the following beneficial effects:

[0073] This invention monitors and smooths network feature vectors in real time, integrates Kalman filtering to smooth buffered data reception speed and accumulated latency, accurately assesses chain reaction risks, and employs dynamic programming to optimize transmission mode switching paths in mixed scenarios. By adjusting buffer filling strategies, triggering adaptation mechanisms, and retransmitting lost packets, this invention ensures smooth playback. Simultaneously, it updates latency records based on feedback data and iteratively optimizes the smoothed network feature vectors, ultimately guaranteeing transmission continuity. Its technical advantages lie in effectively mitigating the risk of interruptions caused by network fluctuations, improving the stability of streaming media services and user experience, and demonstrating strong fault tolerance, particularly in complex network environments. Attached Figure Description

[0074] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0075] Figure 1 This is one of the flowcharts of a fault-tolerant method for streaming media services that supports both pre-broadcast and post-continuous push according to the present invention.

[0076] Figure 2 This is the second flowchart of a fault-tolerant method for streaming media services that supports both pre-broadcast and post-continuous push according to the present invention.

[0077] Figure 3 This is the third flowchart of a fault-tolerant method for streaming media services that supports both pre-broadcast and post-continuous push according to the present invention. Detailed Implementation

[0078] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0079] Please see Figures 1-3 As shown in the figure, this embodiment discloses a fault-tolerant method and system for streaming media services that supports both pre-play and post-push and interrupted push modes. The pre-play mode refers to a transmission mode in which the streaming media client immediately starts decoding and playback after the data volume in the receiving end buffer reaches a preset start threshold, and at the same time dynamically adjusts the push priority of subsequent data segments based on network feedback.

[0080] The interrupted-resume mode refers to a transmission mode that resumes transmission after the network is restored by recording the sequence number of the last successfully received data segment after a connection interruption or persistent timeout is detected.

[0081] Specifically, the methods may include:

[0082] S101. By monitoring network fluctuations and high packet loss rates in real time, the current buffered data receiving speed and end-to-end latency accumulation value are obtained from the streaming media transmission data. The Kalman filter algorithm is used to process these values ​​to filter noise and obtain a smoothed network feature vector.

[0083] The purpose of this step is to eliminate transient noise during the network sampling process and obtain trend data that reflects the true physical state. The specific implementation process is as follows:

[0084] S101-1: By monitoring network fluctuations and high packet loss rates in real time, the system extracts buffered data reception speed and latency accumulation values ​​from streaming media transmission data to obtain initial network status data. Specifically, the system parses control messages from streaming media transmission protocols (such as RTP / RTCP or QUIC) in real time, sampling at preset intervals. (like Extract the current buffer data reception speed (unit: and end-to-end latency accumulation (unit: ), construct the original observation vector .

[0085] S101-2: Based on the initial network state data, the Kalman filter algorithm is used to filter noise from the buffered data receiving speed and accumulated delay value to obtain a smoothed data sequence; specifically, a Kalman filter state model is constructed. The state transition matrix is ​​set. Observation matrix The following formula is used for state prediction and updating:

[0086] Predicted status:

[0087] Calculate the Kalman gain:

[0088] Updated state estimate:

[0089] in, For process noise covariance, To observe the noise covariance, the smooth network feature vector is obtained through iterative calculation. .

[0090] S101-3: For the smoothed data sequence, analyze the changing trend of the smoothed network feature vector to determine whether there are significant fluctuations or delay accumulation phenomena.

[0091] Specifically, targeting In Perform time series analysis and calculate its first derivative to determine the growth slope of time delay accumulation.

[0092] S101-4: If the analysis results show that the smooth network feature vector has significant fluctuations or accumulated delays, prioritize the transmitted data, obtain the processing requirements of key data streams, and determine whether to adjust the transmission strategy.

[0093] Specifically, the system parses the media stream, sets the I-frame data stream to high priority (weight P=1), sets the B-frame or P-frame to low priority (weight P=0.5), and determines whether the current low-priority data is consuming too much bandwidth and the strategy needs to be adjusted.

[0094] S101-5: By assessing the processing requirements of key data streams and combining the results of real-time monitoring, an optimized transmission path is determined, resulting in an adjusted transmission configuration.

[0095] S101-6: Based on the adjusted transmission configuration, update the buffer data receiving speed and latency accumulation control parameters of the streaming media transmission, obtain the latest smooth network feature vector data, and complete the optimization process.

[0096] Specifically, the FEC redundancy control parameter is changed from... Increase to Then proceed to the next sampling cycle to obtain new data. This completes the closed loop of this optimization.

[0097] Specifically, the system first collects real-time RTT and packet loss rate data during streaming media transmission. For example, it calculates the round-trip time (RTT) from RTP packets every 500ms. If the current RTT is 120ms and the average packet loss rate of the last 5 packets is 2.8%, it determines that there is network fluctuation. Next, it obtains the current buffered data reception speed and accumulated latency from the player's buffer API. Assuming the instantaneous buffered data reception speed is 1800kbps and the accumulated latency has reached 320ms, the original buffered data reception speed v_raw=1800 and the accumulated latency d_raw=320 are used as observations and input to the Kalman filter. The Kalman filter uses a one-dimensional state model, and the state vector is defined as... The process noise covariance Q is set as diag(50,120), the observation noise covariance R is set as diag(80,200), the state transition matrix F is an identity matrix, and the observation matrix H is also an identity matrix. The initial state estimate x0=[2000,100]^T and the covariance P0 are diag(400,900). In the k-th iteration, prediction is performed first. Then calculate the Kalman gain. Update the state estimate again Where z_k is the current observation [v_raw, d_raw], after filtering, the smoothed network feature vector is obtained, such as the smoothed buffered data receiving speed. Converging to 1620kbps, latency accumulation The time was reduced to 210ms; finally, an adaptive adjustment was made based on this smoothing value. <1500kbps and If the time exceeds 250ms, the network quality is considered poor, triggering a rate reduction or an increase in redundant packets to achieve a robust response to jitter and packet loss, forming a closed-loop processing link from data acquisition to smooth decision-making.

[0098] in, Let be the state prediction error covariance matrix.

[0099] This is the time delay jitter compensation coefficient. The rate fluctuation tolerance coefficient is preset according to the network environment.

[0100] The minimum theoretical bandwidth required to maintain the current media bitrate.

[0101] S102. Based on the obtained smooth network feature vector, determine whether a chain reaction has occurred. If the accumulated delay exceeds the preset threshold and the buffered data reception speed is lower than the average level, then it is determined that there is a potential interruption risk, and a risk assessment report is generated.

[0102] S102-1: By real-time monitoring of the feature vectors of the smoothing network, dynamic data on latency accumulation and buffered data reception speed are obtained. These data are then preliminarily screened to identify key points of abnormal fluctuations.

[0103] Specifically, the system will Compared with historical average buffered data reception speed (e.g.) (Compare and extract those with a deviation exceeding) The moment is taken as the key point of abnormal fluctuation.

[0104] S102-2: Based on the key points of abnormal fluctuations, analyze the specific distribution of time delay accumulation. If the distribution exceeds the preset threshold, it is determined that there is a potential problem and a corresponding abnormal marker is generated.

[0105] Specifically, acquiring key locations (For example, obtaining) ), set the preset threshold as .because If the distribution exceeds the threshold, the system will generate an error in the data log. Exception markers.

[0106] S102-3: For the generated anomaly markers, combined with the deviation data of the buffered data receiving speed, cross-validation is performed using preset comparison rules to determine whether the speed deviation has a correlation with the accumulation of time delay.

[0107] Specifically, the preset comparison rule is to cross-calculate the risk index. :

[0108]

[0109] Substitute actual data: .

[0110] in, and Weighting coefficients (e.g.) ).like If this is the case, then the speed deviation and time delay accumulation have formed a vicious cycle.

[0111] S102-4: Based on the analysis results of the correlation between speed deviation and time delay accumulation, the triggering conditions for interruption risk are extracted. If the triggering conditions meet the preset standards, a risk warning signal is generated.

[0112] Specifically, the risk warning preset standard is set as follows: Due to the calculation results If the triggering conditions are met, the system generates a risk index. The risk warning signal.

[0113] S102-5: Based on the risk warning signal, call the data analysis module to trace the root cause of potential problems, obtain the main factors affecting network quality, and determine key intervention points.

[0114] Specifically, the module analyzes the jitter variance. If the variance is greater than the bandwidth drop, it confirms that the main factor is "link jitter" rather than "hard congestion," and determines the key intervention point as "expanding the receiver's anti-jitter buffer."

[0115] S102-6: For key intervention points, adjust the priority configuration of transmission paths and obtain optimized network quality data by accumulating status changes through real-time monitoring.

[0116] Specifically, the data packet sending interval is adjusted to be more uniform, and metrics are collected over the next 3 seconds to obtain optimized network quality data.

[0117] S102-7: Based on the optimized network quality data, update the information on the basis of risk assessment, and determine whether there are any new potential problems by continuously tracking changes in indicators.

[0118] Overwrite the newly collected data to From the historical queue, a continuous tracking and evaluation report is generated and output.

[0119] Specifically, in the network quality assessment of streaming media transmission, the system first extracts key indicator data from the monitoring module, such as an accumulated latency value of 280ms and a buffered data reception speed of 1200kbps. These data are then compared with preset thresholds. Assuming the accumulated latency threshold is set at 250ms and the average buffered data reception speed benchmark is 1500kbps, the comparison reveals that the accumulated latency exceeds the threshold by 30ms and the buffered data reception speed is 200kbps below the benchmark. The system automatically determines that the current network status has a potential interruption risk. Next, the system calls the risk assessment algorithm, employing a weighted analysis model, assigning weights of 0.6 to the accumulated latency and 0.4 to the buffered data reception speed, and calculating the risk index. The result was 0.152, exceeding the preset risk threshold of 0.15, confirming that a chain reaction might occur. For further analysis, the system incorporated historical data into the assessment, extracting the average accumulated latency of 230ms and the average buffered data reception speed of 1550kbps over the past 10 minutes. Combined with current data trend analysis, it was found that the accumulated latency was increasing while the buffered data reception speed was continuously decreasing. Based on this, the system generated a detailed risk assessment report, which included quantitative indicators such as a risk index of 0.17, a latency accumulation deviation rate of 12.2%, and a buffered data reception speed deviation rate of 9.7%. Simultaneously, the report data was automatically synchronized to the streaming media scheduling system in conjunction with the business modules to optimize resource allocation, forming a complete logical chain from indicator extraction to risk confirmation and report generation.

[0120] S103. Based on the generated risk assessment report, analyze the transmission mode in the mixed scenario, and use dynamic programming algorithm to optimize the switching path between first-broadcast and second-push and interrupted-continuous-push, so as to obtain the optimized mode switching sequence.

[0121] S103-1: Obtain the accumulated delay value and buffered data reception speed value in the current smooth network feature vector data.

[0122] Specifically, in this step, the current data is directly read from the feature extraction module. and .

[0123] S103-2: By comparing the accumulated delay value with a preset threshold, if the value exceeds the threshold and the buffered data receiving speed is lower than the baseline value, the interruption risk point is determined.

[0124] Specifically, when and At that time, record the current system timestamp as an interruption risk point. .

[0125] S103-3: For points at risk of interruption, obtain the current transmission mode status and determine whether it is a case of "broadcast first, then push" or "continuous push after interruption".

[0126] Specifically, the system reads the player's status register to determine whether it is currently in the "play first, push later" state (State=0) that allows playback to start with missing frames or the "continuous push after interruption" state (State=1) that allows full data transmission to resume. Let's assume that the current state is "play first, push later" (State=0).

[0127] S103-4: Obtain the most recent mode switching history sequence from the transmission mode status.

[0128] Specifically, extract the mode switching logs from the first 10 minutes. For example, the logs show that the percentage of "play first, push later" was 67.2%, and the percentage of "intermittent push" was 32.8%.

[0129] S103-5: Use dynamic programming algorithm to process the historical sequence of mode switching, calculate the total cost of each switching path, and obtain the optimized mode switching sequence corresponding to the path with the lowest cost.

[0130] Specifically, a dynamic programming algorithm is used to calculate the switching path. The state transition cost is defined. Signaling overhead and transmission loss caused by mode switching For pattern Predict the probability of stuttering. Calculate the optimal cost:

[0131]

[0132] By backtracking, the path with the lowest cost is obtained, which is the optimal mode switching control sequence.

[0133] S103-6: Perform dynamic switching operation of transmission mode according to the optimization mode switching sequence, and update the transmission mode status.

[0134] Specifically, the network socket is disconnected and reconnected, and the status register is updated. (Continuous push).

[0135] S103-7: By dynamically switching the transmission mode status after the operation, obtain new smooth network feature vector data and determine whether the interruption risk points have disappeared.

[0136] Specifically, in the new model for collecting RTT, if the newly calculated risk index If so, it is confirmed that the point of interruption risk has disappeared.

[0137] Specifically, in the process of generating a risk assessment report for a hybrid transmission scenario, the system first extracts relevant indicators of the current transmission mode from the real-time monitoring data stream. For example, the cumulative jitter value in the current broadcast-then-push mode is 420ms, and the network switching delay is 180ms. Then, a dynamic programming algorithm is used to optimize the path for both modes. The state transition cost function is set as a weighted sum of jitter increment and switching delay, with a jitter weight of 0.65 and a switching delay weight of 0.35. The algorithm uses 120 sampling points collected in the past 5 minutes as the observation sequence. Through forward recursion, the optimal cost of switching from the current broadcast-then-push to continuous push is calculated to be 0.191, while the cumulative cost of maintaining broadcast-then-push is 0.238. The system determines that the switching path cost is lower, but it still needs to be judged in conjunction with a risk threshold. Next, the system loads a preset interruption risk threshold of 0.22. After comparing the optimal switching cost of 0.238 with the threshold, it is confirmed that it exceeds 0.018, triggering a chain reaction warning. At the same time, the predicted interruption probability within the next 30 seconds under the current mode is calculated to be 31.4%. To deepen the analysis, the system introduced historical pattern sequence data, extracting the proportion of the first-play-then-push mode (67.2%) and the discontinuous-push mode (32.8%) within the first 10 minutes. Using exponential smoothing, the system predicted the pattern distribution for the next cycle, finding that without intervention, the proportion of the first-play-then-push mode would further increase to 72.6%, exacerbating the risk of jitter accumulation. Based on this, a risk assessment report was generated. The report detailed the optimal switching sequence as first-play-then-push - brief buffer adjustment - discontinuous-push, with expected jitter values ​​of 380ms, 210ms, and 145ms for each stage, a risk quantification score of 0.238, an interruption probability of 31.4%, and a mode switching benefit improvement rate of 18.7%. The report results were automatically pushed to the transmission scheduling engine to trigger mode adjustment decisions, forming a complete automated closed-loop process from indicator extraction, path optimization, risk confirmation to report output.

[0138] S104. Based on the obtained mode switching sequence, trigger the activation condition of the adaptation mechanism. If the smoothing network feature vector shows continuous fluctuations, adjust the buffer filling strategy to maintain smooth playback.

[0139] S104-1: By using the mode switching sequence, obtain the current network quality monitoring data, determine whether the trigger conditions are met based on the fluctuation of the indicators, and obtain a preliminary judgment on the response requirements.

[0140] Specifically, the system analyzes the parameters of the newly switched transmission mode and extracts the current smoothed buffer data reception speed. In recent sampling points (e.g.) Standard deviation .like Therefore, it can be preliminarily determined that there is a need for fluctuation response.

[0141] S104-2: If the preliminary response to the demand indicates that there are fluctuations in network quality indicators, then extract the specific fluctuation frequency and amplitude from the monitoring data, compare them with the preset threshold, and determine whether to activate the adaptation mechanism.

[0142] Specifically, further calculate the latency jitter amplitude. and fluctuation frequency The preset activation threshold is... and If the conditions are met, the adaptation mechanism is formally activated.

[0143] S104-3: Based on the activation state of the adaptation mechanism, obtain the current buffer filling level. In the case of insufficient buffer filling, calculate the filling rate adjustment parameter to obtain the direction of strategy adjustment.

[0144] Specifically, obtain the current actual buffer duration. and basic buffer duration (like ). Calculate the target buffer duration based on jitter characteristics: .like If so, the strategy adjustment direction is "accelerate filling".

[0145] S104-4: By adjusting the direction of the strategy, dynamically respond to the actual needs of buffer filling, adjust the filling rate and priority allocation, and determine the data replenishment rhythm of the buffer.

[0146] Specifically, an acceleration command is sent to the transmission control layer to set a new target pull rate. It also prioritizes the application layer reception of media stream data packets to the highest level.

[0147] S104-5: Based on the data replenishment rhythm of the buffer, continuously monitor changes in network quality, track the persistence of indicator fluctuations in real time, and obtain the status of ensuring playback continuity.

[0148] Specifically, it updates every 200ms. The value is used to calculate the difference between the buffer consumption rate and the fill rate. If the fill rate is greater than the consumption rate for three consecutive cycles, then the playback continuity is confirmed to be in a safe and secure state.

[0149] S104-6: By playing the continuous guarantee status and combining the results of quality monitoring, dynamically respond to whether there is a decline in network quality, and determine whether to maintain the current strategy adjustment plan.

[0150] Specifically, if at this time the network packet loss rate suddenly increases, leading to If a sudden drop in flow rate disrupts the protection mechanism, the system will dynamically adjust the upper limit of the flow rate to prevent the TCP congestion control window from collapsing.

[0151] Based on the maintenance status of the strategy adjustment plan, obtain the latest switching sequence data, record the frequency and effect of mode switching, and determine the basis for subsequent dynamic response optimization.

[0152] Specifically, this The adjustment amount and corresponding lag avoidance duration are written to the local log database, and the historical strategy effectiveness weights are updated using the Exponential Weighted Moving Average (EWMA) algorithm for reference in the next fluctuation.

[0153] Specifically, the system updates the underlying buffer baseline to If the current actual buffer duration Then, an acceleration request is sent to the network transmission control layer (e.g., by increasing the application layer's pull priority or increasing the TCP receive window) to force a higher fill rate.

[0154] Specifically, in mixed transmission scenarios, the system automatically triggers the activation conditions of the adaptation mechanism through the generated mode switching sequence. First, it monitors the smooth network feature vector in real time, collecting network fluctuation data over the past 3 minutes. It finds an average latency jitter of 350ms and a packet loss rate of 2.3%, comparing these indicators with preset fluctuation thresholds (300ms jitter, 1.5% packet loss rate) to confirm that the current network quality is continuously fluctuating, meeting the activation conditions. Next, the system calls a preset buffer filling strategy adjustment algorithm. Based on the current playback rate of 1.2Mbps and a buffer occupancy rate of 45%, it calculates that the buffer filling target needs to be increased from the default 5 seconds to 8 seconds to enhance resistance to fluctuations. Simultaneously, it uses linear regression analysis to predict the network jitter trend over the next 2 minutes, finding a 62.8% probability that the jitter may further increase to 400ms. Based on this, it dynamically adjusts the filling rate to 1.5Mbps to ensure the buffer reaches the target value within 15 seconds. Subsequently, the system, combined with a playback smoothness evaluation model, compared the current buffer state with historical data. Analysis showed that the probability of smoothness interruption decreased from 28.5% to 16.7% after the adjustment. An adjustment log was automatically generated, recording key indicators before and after the filling strategy change, such as jitter decreasing from 350ms to the expected 320ms and buffer occupancy increasing from 45% to 68%, providing data support for subsequent optimization. Finally, the system synchronized the adjusted strategy parameters to the transmission control module to ensure real-time application of the new strategy during playback. Simultaneously, the system assessed the impact on user experience from related business modules, calculating an expected 3.2 percentage point increase in user satisfaction score after the adjustment, thus forming a complete automated process from network monitoring to strategy adjustment and effect evaluation.

[0155] S105. Based on the adjusted buffer filling strategy, obtain feedback data on reliable transmission, determine whether lost packets need to be retransmitted, and if the feedback data indicates a high packet loss rate, perform a retransmission operation and update the latency accumulation record.

[0156] S105-1: Based on the adjusted buffer filling strategy, obtain reliable transmission feedback data, determine whether the packet loss rate exceeds the preset threshold, and if the packet loss rate is high, mark the lost data packets and trigger a retransmission request to obtain the retransmission queue record.

[0157] Specifically, the system listens for RTCPNACK or ACK packets from underlying protocols (such as QUIC). It then calculates the packet loss rate within the sliding window. .like If the sequence number is not confirmed, the data packet will be pushed into the retransmission queue.

[0158] S105-2: Obtain the retransmission delay data of each lost packet through the retransmission queue record, calculate the cumulative delay increment, and determine the current playback delay status.

[0159] Specifically, the round-trip time of each retransmitted packet is calculated. Since retransmission inevitably increases physical latency, the cumulative latency increment is calculated. This increment is then summed to determine the current physical delay state.

[0160] S105-3: Based on the current playback delay status, obtain the video frame decoding timestamp difference to determine whether decoding starvation has occurred. If decoding starvation occurs, adjust the data packet decoding priority to obtain an emergency frame replacement sequence.

[0161] Specifically, this is the core decision in this step: Extract the expected decoding timestamp of the frame to which the lost packet belongs. Get the current local playback clock. Calculate and decode the hunger risk value .like If decoding starvation is detected, an emergency frame interpolation sequence is generated. This indicates that the decoding has been missed, and the packet will be discarded without being retransmitted.

[0162] S105-4: By using emergency frame interpolation sequences, the continuity of sequence numbers of received data packets is detected, the position of sequence number gaps is extracted, and the list of actual gap positions is determined.

[0163] Specifically, iterate through the received RTP sequence numbers in the buffer (e.g., to ), find missing indexes (e.g. , This generates a precise list of gap locations.

[0164] S105-5: Based on the actual gap location list, filter out the key frame packets that have not yet arrived from the reliable transmission feedback data, mark them as high-priority retransmission targets, and obtain the key retransmission subset.

[0165] Specifically, the data types corresponding to the missing sequence numbers are parsed, and only data packets belonging to I-frames or base layer P-frames are marked as high-priority retransmission targets. Discardable B-frames are stripped to form a critical retransmission subset.

[0166] S105-6: By using the key retransmission subset, dynamically update the retransmission window size and retransmission timer duration to determine the new round of retransmission rhythm.

[0167] Specifically, the backoff base of the retransmission timer (RTO) will be changed from the current... shortened to A dedicated emergency sending window is also opened (with priority over regular data push).

[0168] S105-7: Based on the new round of retransmission rhythm, continuously obtain the latest reliable transmission feedback data, update the cumulative delay increment and packet loss rate statistics in real time, and obtain the transmission quality assessment results.

[0169] Specifically, after the retransmission is completed, an actual arrival confirmation is obtained, and the final [data / information] is [acknowledged / determined]. Superimpose the accumulated end-to-end delay onto the baseline delay and update the system's end-to-end delay record. ( For example, a smoothing decay factor. ).

[0170] Specifically, in a mixed transmission scenario, the system continuously collects feedback data from the reliable transmission module based on the adjusted buffer filling strategy, and obtains the acknowledgment status of each data packet in the past 60 seconds in real time. It finds that a total of 17 packets have not received ACK acknowledgments. Combined with the current packet loss rate reaching 3.7%, exceeding the preset retransmission trigger threshold of 2.0%, the retransmission judgment process is immediately initiated. The system calls a sliding window-based packet loss detection algorithm to check the status of packets with sequence numbers 4521 to 4580 one by one. It confirms that 9 packets are continuously lost, and another 8 are intermittent single-packet losses. Subsequently, the system prioritizes retransmitting the 9 continuously lost packets with sequence numbers 4523, 4527, 4534, 4541, 4548, 4552, 4559, 4563, and 4571 through a fast retransmission mechanism, and simultaneously records the sending timestamps of these packets in the latency accumulation table. The retransmission employs an exponential backoff strategy, with the initial retransmission interval set to 1.5 times the current RTT, approximately 180ms. If no acknowledgment is received within 2 seconds, the interval doubles to 360ms for a second retransmission. Simultaneously, the system updates the latency accumulation record, calculating the additional latency increment introduced by this retransmission as an average of 42ms, and weighted averages it with the historical latency accumulation values ​​from the previous 5 minutes, resulting in a current total latency accumulation of 217ms, close to the warning threshold of 250ms. Next, a Kalman filter algorithm is used to predict the packet loss rate for the next 30 seconds. Inputting the most recently collected packet loss sequence data, the filtered packet loss rate is expected to stabilize at 2.9%. The packet loss rate shows a significant downward trend after retransmission; therefore, it is decided not to expand the retransmission window range for the time being, but only to selectively retransmit high-priority video keyframe packets. After all retransmission operations are completed, the system will feed back the updated confirmation packet ratio to the transmission control module, recording a retransmission success rate of 94.1%. It will also store the accumulated latency records in association with changes in packet loss rate, providing a basis for subsequent adaptive adjustment of congestion control parameters. The entire process is executed continuously in the background by an automated script, ensuring that playback continuity is not significantly affected.

[0171] S106. By updating the accumulated time delay records, determine the degree of mitigation of the chain reaction, and recalculate the smoothing network feature vector using the Kalman filter algorithm to obtain the corrected index value.

[0172] S106-1: By accumulating and recording latency, extract latency change data in network transmission, classify the fluctuation of latency changes, and obtain a preliminary fluctuation distribution state.

[0173] Specifically, extract the updates from the preceding steps. Sequence. The delay variation is decomposed into "queue delay" (which grows linearly) and "transmission jitter" (which follows a Gaussian distribution), quantifying the fluctuation distribution of the system.

[0174] S106-2: Based on the preliminary fluctuation distribution, analyze the correlation between network transmission and data stability. If the fluctuation distribution exceeds the preset threshold, mark the network status and determine the abnormal fluctuation range.

[0175] Specifically, calculate the latest risk index. .like If the value is still higher than the preset alarm threshold, then the time period is marked as an abnormal fluctuation range.

[0176] S106-3: Obtain quality index data under network conditions through abnormal fluctuation ranges, and use Kalman filtering algorithm to smooth the quality index to obtain corrected evaluation results.

[0177] Specifically, this step is the core of fault-tolerant feedback. It involves calculating the chain reaction mitigation index. The new data observed Re-enter the Kalman filter state model from step S101 and perform prediction and update.

[0178] S106-4: Based on the revised evaluation results, update the frequency of indicator updates, monitor the dynamic changes in transmission performance in real time, and determine the current stability level of network transmission.

[0179] Specifically, if the degree of relief (The relief is not significant), the system forcibly reduces the sampling and update frequency of the Kalman filter from... Upgraded to This is to capture high-frequency jitter and set the current stability level to "vulnerable".

[0180] S106-5: Analyze the degree of matching between delay changes and data stability through stability levels. If the degree of matching is lower than the preset threshold, adjust the sampling interval of dynamic monitoring to obtain an optimized monitoring rhythm.

[0181] Specifically, if the level is "fragile" and the data variance and delay increment are severely mismatched, the value of the process noise covariance Q is dynamically reduced (to trust the observation data more), while maintaining a high-frequency sampling interval.

[0182] S106-6: Based on the optimized monitoring schedule, continuously acquire the latest network status data, compare and analyze the results with the evaluation results, and determine the long-term fluctuation trend of network transmission.

[0183] Specifically, after a smoothing period (such as...) After that, output the corrected smoothing index value that has withstood two tests. .

[0184] Specifically, in a hybrid transmission scenario, the system continuously monitors the network status based on the latency data accumulated since the previous retransmission. By reading the latest total latency value of 217ms from the latency accumulation table and combining it with the round-trip latency samples of each retransmitted packet within the last 300 seconds, the system calculates that the current cascading effect has been alleviated, with the latency increment decreasing from 42ms to an average of 19ms, indicating that the network congestion pressure has been initially relieved. The system then initiates a Kalman filter algorithm to smooth the network feature vector and re-corrects it. The process noise covariance Q is set to 0.0004, the measurement noise covariance R to 0.0025, and the state transition matrix adopts a first-order autoregressive model. The previous time-series filtered estimate is used as a priori, and the input is the most recently collected 30 RTT sample sequences (with an average RTT fluctuation range between 112ms and 148ms). After prediction-update iteration, the corrected smoothed RTT value is 131.7ms, the corrected packet loss rate estimate is 2.4%, and the jitter standard deviation converges to 7.8ms. Based on a comparison of the corrected metrics with historical thresholds, the system found that the current RTT was 150ms below the warning value and the packet loss rate had fallen below the threshold of 2.5%, indicating that the chain reaction had been mitigated to a safe range. Therefore, the system lowered the retransmission priority and adjusted the congestion window growth factor from a linear increase of 0.8 to a conservative increase of 0.4. Simultaneously, the retransmission probability of non-critical frame packets in the video stream was reduced from 0.65 to 0.25 to minimize the additional latency overhead caused by unnecessary retransmissions. The system also weighted and fused the corrected metrics with the results of the previous two filtering iterations, with weights of 0.5, 0.3, and 0.2, respectively, calculating a comprehensive network health index of 0.87, close to the ideal value of 0.90. This index was written to the transmission control parameter table in real time, providing a basis for adaptive bitrate decisions in the next cycle. The entire analysis and adjustment process was automatically triggered by a background timer task every 15 seconds, ensuring that the transmission strategy dynamically matched the actual network condition.

[0185] S107. Based on the corrected index values, optimize the transmission mode in mixed scenarios to ensure continuity, determine whether playback smoothness has been restored, and if not, repeatedly trigger the adaptation mechanism until reliable transmission is stable.

[0186] S107-1: Obtain the corrected index value and the current mixed scene parameters.

[0187] Specifically, extract from the filter output. It also obtains the current media bitrate, resolution, and other mixed scene parameters.

[0188] S107-2: Determine the transmission mode switching strategy based on the corrected index value and mixed scenario parameters.

[0189] Specifically, the modified indicators are substituted into the dynamic programming model of S103 to reassess whether the current "discontinuous push" or "broadcast first, push later" mode is still the minimum cost path.

[0190] S107-3: Path switching is performed using a transmission mode switching strategy to obtain the delay sequence after switching.

[0191] Specifically, if the current cost is not optimal, a reverse mode switch is performed (e.g., switching back from "disconnected push" to "broadcast first, then push" after network recovery), and RTT data within 3 seconds after the switch is collected to form a delay sequence.

[0192] S107-4: The delayed sequence is processed by the Kalman filter algorithm to obtain the smoothing quality estimate.

[0193] Specifically, the steady-state convergence value of the above sequence is calculated using a Kalman filter to obtain the final smoothing quality estimate.

[0194] S107-5: Determine the smoothness of playback based on the smoothness quality estimate.

[0195] Specifically, define the playback stability index. As a rigid criterion for judgment.

[0196]

[0197] in .Compare With stability threshold (e.g.) ).like If so, it is determined that playback has been restored smoothly.

[0198] S107-6: If the playback is not stable, the transmission mode switching strategy will be redefined.

[0199] Specifically, if This indicates that the physical network or buffer is still not out of danger, and the system locks the current bitrate increase operation to prevent further congestion.

[0200] S107-7: Based on the smoothness quality estimate and playback smoothness status, continuously monitor the transmission path changes until a reliable and stable state is achieved.

[0201] Specifically, the system will currently Feedback is sent to the control input, triggering a re-execution of the buffer adjustment in S104 and the emergency retransmission logic in S105, forming a closed-loop control loop until... continued After more than one cycle, reliable and stable transmission is confirmed.

[0202] Specifically, in a mixed transmission scenario, the system continuously collects the time series of the current video stream's segment reception completion. By querying the latest batch of 30 consecutive segments in the transmission log database, it was found that the average reception interval has shortened from the initial 48ms to the current 21ms, while the maximum interval has decreased from the previous peak of 91ms to 37ms. This indicates that the playback buffer feeding rate has stabilized, and the risk of continuous interruption has been initially determined to be reduced. The system then uses an exponentially weighted moving average algorithm to smoothly evaluate the buffer health, setting the smoothing factor α to 0.15. Using the most recently collected 60-second sample of the buffer's remaining duration as the input sequence (the remaining duration sequence ranges from 1.2 seconds to 4.8 seconds), the current smoothed buffer duration is calculated to be 3.1 seconds, exceeding the target minimum safety threshold of 2.5 seconds. Based on this result, the system further combines the real-time bandwidth estimate with Kalman filtering to dynamically track the available bandwidth. The process noise covariance Q is set to 0.001, and the measurement noise covariance R is set to 0.005. A constant-rate model is used for state transition, and the prior bandwidth is taken from the previous cycle's filtered value. After multiple prediction and correction iterations, the corrected bandwidth estimate converges to 4.82 Mbps, a decrease of only 5.5% compared to the historical peak of 5.1 Mbps. The system inputs the smoothing buffer duration of 3.1 seconds and the corrected bandwidth of 4.82 Mbps into the decision tree model for a comprehensive judgment on smoothness. The model rule is set so that when the buffer duration is greater than 2.5 seconds and the bandwidth estimate is greater than 4.5 Mbps, it is determined that playback smoothness has been restored. The current condition is met, so the adaptive bitrate is increased from 720p@1800kbps to 1080p@3200kbps. At the same time, the keyframe interval control parameter is adjusted from the original 2 seconds to 1.5 seconds to improve the continuity of the picture, and the FEC redundancy ratio is reduced from 12% to 6% to optimize bandwidth utilization. The system also performs exponential decay fusion of the current judgment result with the previous two decision records, assigning a weight of 0.6 to the latest record, 0.25 to the previous one, and 0.15 to the one before that. The calculated playback stability index is 0.91, which exceeds the recovery threshold of 0.85. The system then immediately updates the playback control configuration file for the next round of bitrate adaptation and buffer scheduling modules to read. The entire monitoring, evaluation, and optimization process is automatically executed by the real-time daemon process every 8 seconds until the system confirms that the transmission link is stable in the long term.

[0203] This invention provides a fault-tolerant streaming media service system that supports both pre-broadcast and post-continuous push, mainly comprising:

[0204] The real-time monitoring and data acquisition module is used to obtain the current buffered data reception speed and latency accumulation value from streaming media transmission data by monitoring network fluctuations and high packet loss rates in real time.

[0205] The smoothing network feature vector processing module is used to process these values ​​using the Kalman filter algorithm to filter out noise and obtain smoothed network feature vectors.

[0206] The chain reaction judgment and risk assessment module is used to determine whether a chain reaction has occurred based on the obtained smooth network feature vector. If the accumulated delay exceeds the preset threshold and the buffered data reception speed is lower than the average level, it is determined that there is a potential interruption risk, and then a risk assessment report is generated.

[0207] The transmission mode analysis and optimization module is used to analyze the transmission mode in a mixed scenario based on the generated risk assessment report. It uses a dynamic programming algorithm to optimize the switching path between pre-broadcast and post-continuous push, and obtains an optimized mode switching sequence.

[0208] The adaptation mechanism triggering and buffer adjustment module is used to trigger the activation condition of the adaptation mechanism based on the obtained mode switching sequence. If the smoothing network feature vector shows continuous fluctuations, the buffer filling strategy is adjusted to maintain smooth playback.

[0209] The feedback data processing and retransmission module is used to obtain reliable transmission feedback data based on the adjusted buffer filling strategy, determine whether lost packets need to be retransmitted, and if the feedback data indicates a high packet loss rate, then a retransmission operation is performed and the latency accumulation record is updated.

[0210] The chain reaction mitigation and index correction module is used to determine the degree of chain reaction mitigation by updating the accumulated latency records. It uses the Kalman filter algorithm to recalculate the smoothing network feature vector to obtain the corrected index value. Based on the corrected index value, it optimizes the transmission mode in the mixed scene to ensure continuity and determines whether playback smoothness has been restored. If it has not been restored, it triggers the adaptation mechanism repeatedly until reliable transmission is stable.

[0211] It should be noted that all preset thresholds, benchmark values, and weight parameters involved in the above-mentioned scheme of this invention (such as preset start-up thresholds, latency warning thresholds, historical average buffer data reception speed benchmarks, preset trigger packet loss rate thresholds, and weight coefficients in each evaluation model) are specifically determined based on statistical analysis and machine learning model training of a limited number of historical actual transmission data. Specifically, the system collects at least 10,000 streaming media transmission log data under different network jitter conditions during the offline phase, extracts latency, packet loss, and buffer speed features to form a sample set; then, gradient descent or grid search is used to iteratively optimize each preset parameter, and convergence is determined by the objective function (such as setting the stuttering rate to be less than 1% and the bandwidth utilization to be the highest); when the objective function converges, the parameter set at this time is recorded as the final preset parameters and applied to the online fault-tolerant system. For example, regarding the latency warning threshold, the system statistically analyzed playback interruption samples under historical weak network conditions. After data fitting, it was found that when the end-to-end latency accumulation value exceeds 250ms, the probability of subsequent severe stuttering increases exponentially to over 85%. Therefore, the system sets the latency warning threshold to 250ms.

[0212] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, the personal information processing rules are clearly informed through signs / information, and authorization is obtained through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0213] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push, characterized in that, include: By monitoring network fluctuations and high packet loss rates in real time, the current buffer data receiving speed and end-to-end latency accumulation value are obtained from streaming media transmission data according to a preset sampling period. The original observation vector is constructed, and the Kalman filter algorithm is used to calculate the Kalman gain in combination with the set state transition matrix and observation matrix for state prediction and update. These values ​​are then processed to filter noise and obtain a smoothed network feature vector. Based on the obtained smooth network feature vector, a risk index composed of delay accumulation deviation and buffer data reception speed deviation is calculated. If the risk index is greater than a preset threshold and the buffer data reception speed is lower than the average level, a potential interruption risk is determined, and a risk warning signal containing the risk index is generated. For the generated risk warning signal, the transmission mode status of the client player's status register, which belongs to the first-play-then-push or the interrupted-continuous-push mode, is obtained. The state transition cost and transmission loss are calculated by dynamic programming algorithm, so as to optimize the switching path of first-play-then-push and interrupted-continuous-push and obtain the optimized mode switching sequence corresponding to the lowest cost path. Based on the obtained mode switching sequence, the specific latency jitter amplitude and fluctuation frequency are extracted. If the smoothed network feature vector shows continuous fluctuation, the target buffer duration is calculated based on the latency jitter amplitude and the smoothed network feature vector, and the corresponding pull rate target is sent to the transmission control layer to maintain smooth playback. Based on the adjusted target buffer duration and pull rate target, obtain reliable transmission feedback data, determine whether lost packets need to be retransmitted, if the feedback data indicates a high packet loss rate, perform a retransmission operation, calculate the round-trip time of the retransmitted packets to obtain the cumulative delay increment, and add the cumulative delay increment to the base delay to update the delay accumulation record. By updating the accumulated delay records, the magnitude of the delay increment reduction caused by retransmission is determined as the mitigation level. The Kalman filter algorithm is used to recalculate the smoothing network feature vector to obtain the corrected index value. If the mitigation level is lower than the preset value, the sampling and update frequency of the Kalman filter is forcibly increased. Based on the corrected index values, the dynamic programming algorithm is re-executed to optimize the transmission mode to ensure continuity. The playback stability index is calculated by combining the corrected index values ​​with the actual buffer duration. Based on this index, it is determined whether playback smoothness has been restored. If not, the target step of adjusting the streaming rate is triggered repeatedly until the playback stability index reaches the stable threshold.

2. The fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push as described in claim 1, characterized in that, The process involves real-time monitoring of network fluctuations and high packet loss rates. Following a preset sampling period, the current buffered data reception speed and end-to-end latency accumulation are obtained from the streaming media transmission data to construct an original observation vector. A Kalman filter algorithm is then used, combined with a predefined state transition matrix and observation matrix, to calculate the Kalman gain for state prediction and updating. These values ​​are then processed to filter noise, resulting in a smoothed network feature vector, including: By monitoring network fluctuations and high packet loss rates in real time, the buffered data reception speed and latency accumulation value are extracted from the streaming media transmission data to obtain the initial network status data. Based on the initial network state data, the Kalman filter algorithm is used to filter noise from the buffered data receiving speed and the accumulated delay value to obtain a smoothed data sequence. For the smoothed data sequence, analyze the changing trend of the smoothing network feature vector to determine whether there are significant fluctuations or delay accumulation phenomena. If the analysis results show that the smooth network feature vector has significant fluctuations or accumulated delays, the transmitted data is prioritized to obtain the processing requirements of key data streams and determine whether to adjust the transmission strategy. By assessing the processing requirements of key data streams and combining the results of real-time monitoring, an optimized transmission path is determined, resulting in an adjusted transmission configuration. Based on the adjusted transmission configuration, update the buffer data receiving speed and forward error correction redundancy control parameters for streaming media transmission, obtain the latest smoothing network feature vector data, and complete the optimization process.

3. The fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push as described in claim 1, characterized in that, The process involves calculating a risk index based on the obtained smooth network feature vector, composed of the delay accumulation deviation and the buffered data reception speed deviation. If the risk index is greater than a preset threshold and the buffered data reception speed is lower than the average level, a potential interruption risk is identified, and a risk warning signal containing the risk index is generated, including: By monitoring the feature vectors of the smoothing network in real time, dynamic data on latency accumulation and buffered data reception speed are obtained. These data are then used for preliminary screening to identify key points of abnormal fluctuations. Based on the key points of abnormal fluctuations, the accumulated time delay value of the point is extracted and compared with a preset threshold. If the accumulated time delay value exceeds the preset threshold, a potential problem is identified, and a corresponding abnormality marker is generated. For the generated anomaly markers, the deviation data, which is composed of the difference between the actual rate of buffer data reception and the historical average, is cross-validated using a preset comparison rule to determine whether the speed deviation is related to the accumulation of time delay. Based on the analysis of the correlation between speed deviation and time delay accumulation, the triggering conditions for interruption risk are extracted. If the triggering conditions meet the preset standards, a risk warning signal is generated. Based on the risk warning signal, the data analysis module is invoked to trace the root cause of potential problems, identify the main factors affecting network quality, and determine key intervention points; For key intervention points, the priority configuration of transmission paths is adjusted, and optimized network quality data is obtained by accumulating status changes through real-time monitoring. Based on the optimized network quality data, the information on risk assessment is updated, and by continuously tracking changes in indicators, it is determined whether there are any new potential problems.

4. The fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push as described in claim 1, characterized in that, In response to the generated risk warning signal, the system obtains the transmission mode status (play-before-push or discontinuous push) from the client player's status register. A dynamic programming algorithm is then used to calculate the state transition cost and transmission loss to optimize the switching path between play-before-push and discontinuous push, resulting in the optimized mode switching sequence corresponding to the lowest-cost path. This sequence includes: Obtain the accumulated time delay value and the buffered data reception speed value from the current smooth network feature vector data; By comparing the accumulated delay value with a preset threshold, if the value exceeds the threshold and the buffered data reception speed is lower than the benchmark value, the interruption risk point is determined. For points at risk of interruption, obtain the current transmission mode status and determine whether it is a case of broadcast first and then push or interruption followed by continuous push; Obtain the most recent mode switching history sequence from the transmission mode status; A dynamic programming algorithm is used to process the historical sequence of mode switching, calculate the total cost of state transition cost and transmission loss for each switching path, and obtain the optimized mode switching sequence corresponding to the path with the lowest cost. Perform dynamic switching operations on the transmission mode according to the optimization mode switching sequence, and update the transmission mode status; By dynamically switching the transmission mode status, new smooth network feature vector data is obtained to determine whether the interruption risk points have disappeared.

5. A fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push as described in claim 1, characterized in that, The process involves extracting the specific latency jitter amplitude and fluctuation frequency from the obtained mode switching sequence. If the smoothed network feature vector shows continuous fluctuations, the target buffer duration is calculated based on the latency jitter amplitude and the smoothed network feature vector. A corresponding pull rate target is then sent to the transmission control layer to maintain smooth playback. This includes: By using the mode switching sequence, monitoring data on the current network quality is obtained. Based on the fluctuation of the indicators, it is determined whether the triggering conditions are met, and a preliminary judgment on the response requirements is obtained. If the initial response to the demand indicates that there are fluctuations in network quality indicators, then the specific fluctuation frequency and amplitude are extracted from the monitoring data, and a preset threshold is used for comparison to determine whether to activate the adaptation mechanism. Based on the activation status of the adaptation mechanism, the current buffer filling level is obtained. In the case of insufficient buffer filling, the filling rate adjustment parameter is calculated to obtain the direction of strategy adjustment. By adjusting the direction of the strategy, we can dynamically respond to the actual needs of buffer filling, adjust the filling rate and priority allocation, and determine the data replenishment rhythm of the buffer. Based on the data replenishment rhythm of the buffer, we continuously monitor changes in network quality and track the persistence of indicator fluctuations in real time to ensure the continuity of playback. By playing the continuous guarantee status and combining the results of quality monitoring, we can dynamically respond to whether there is a decline in network quality and determine whether to maintain the current strategy adjustment plan. Based on the maintenance status of the strategy adjustment plan, obtain the latest switching sequence data, record the frequency and effect of mode switching, and determine the basis for subsequent dynamic response optimization.

6. A fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push as described in claim 1, characterized in that, The process involves obtaining reliable transmission feedback data based on the adjusted target buffer duration and target pull rate, determining whether lost packets need to be retransmitted, and performing a retransmission operation if the feedback data indicates a high packet loss rate. The round-trip time of the retransmitted packets is then calculated to obtain the cumulative delay increment. This cumulative delay increment is then added to the base delay to update the delay accumulation record. This includes: Based on the adjusted buffer filling strategy, reliable transmission feedback data is obtained to determine whether the packet loss rate exceeds the preset threshold. If the packet loss rate is high, lost data packets are marked and retransmission requests are triggered to obtain retransmission queue records. By recording the retransmission queue, the retransmission delay data of each lost packet is obtained, the cumulative delay increment is calculated, and the current playback delay status is determined. Based on the current playback delay status, obtain the video frame decoding timestamp difference to determine whether decoding starvation has occurred. If decoding starvation has occurred, adjust the data packet decoding priority to obtain an emergency frame replacement sequence. By using emergency frame interpolation sequences, the continuity of sequence numbers of received data packets is detected, the position of sequence number gaps is extracted, and the list of actual gap positions is determined. Based on the list of actual gap locations, filter out the data packets belonging to I-frames or base layer P-frames that have not yet arrived from the reliable transmission feedback data, mark them as high-priority retransmission targets, and obtain the critical retransmission subset; By dynamically updating the retransmission window size and retransmission timer duration using key retransmission subsets, a new round of retransmission rhythm is determined. Based on the new round of retransmission schedule, we continuously acquire the latest reliable transmission feedback data, update the cumulative latency increment and packet loss rate statistics in real time, and obtain the transmission quality assessment results.

7. A fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push according to claim 1, characterized in that, The reduction in latency increment caused by retransmission is determined by updating the latency accumulation record as the mitigation level. The Kalman filter algorithm is then used to recalculate the smoothing network feature vector to obtain a corrected index value. If the mitigation level is lower than a preset value, the sampling and update frequency of the Kalman filter is forcibly increased, including: By accumulating and recording latency data, latency variation data in network transmission is extracted. Queuing latency and transmission jitter are classified and processed according to the fluctuation of latency variation to obtain a preliminary fluctuation distribution status. Based on the preliminary fluctuation distribution, the correlation between network transmission and data stability is analyzed. If the fluctuation distribution exceeds the preset threshold, the network status is marked to determine the abnormal fluctuation range. By identifying abnormal fluctuation ranges, quality index data under network conditions is obtained. The Kalman filter algorithm is then used to smooth the quality indexes, resulting in a corrected evaluation result. Based on the revised assessment results, the frequency of indicator updates is adjusted, and real-time monitoring of dynamic changes in transmission performance is conducted to determine the current stability level of network transmission. By analyzing the degree of matching between delay variation and data stability through stability level, if the degree of matching is lower than the preset threshold, the value of process noise covariance Q is dynamically reduced to obtain the optimized monitoring rhythm. Based on the optimized monitoring schedule, the latest network status data is continuously acquired, and compared and analyzed with the evaluation results to determine the long-term fluctuation trend of network transmission.

8. A fault-tolerant method for streaming media services supporting both pre-broadcast and intermittent / continuous push as described in claim 1, characterized in that, Based on the revised index values, the dynamic programming algorithm is re-executed to optimize the transmission mode to ensure continuity. A playback stability index is calculated by combining the revised index values ​​with the actual buffer duration. This index is used to determine whether playback smoothness has been restored. If not, the target step of adjusting the streaming rate is repeatedly triggered until the playback stability index reaches a stable threshold, including: Obtain the corrected index value and the current mixed scene parameters; Based on the corrected index values ​​and mixed scenario parameters, determine the transmission mode switching strategy; A path switching strategy is used to perform path switching, resulting in a delayed sequence after the switch. The delayed sequence is processed by the Kalman filter algorithm to obtain a smoothing quality estimate. The playback smoothness status is determined by calculating the playback stability index based on the smoothness quality estimate and the actual buffer duration. If the playback stability index does not reach the stability threshold, the transmission mode switching strategy will be redefined. Based on the smoothness quality estimate and playback smoothness status, the transmission path changes are continuously monitored until the playback stability index reaches a stable threshold.

9. A fault-tolerant system for streaming media services that supports both pre-broadcast and post-continuous streaming, characterized in that, The system includes: The real-time monitoring and data acquisition module is used to obtain the current buffered data reception speed and latency accumulation value from streaming media transmission data by monitoring network fluctuations and high packet loss rates in real time. The smoothed network feature vector processing module is used to monitor network fluctuations and high packet loss rates in real time, obtain the current buffer data receiving speed and end-to-end latency accumulation value from streaming media transmission data according to a preset sampling period, construct the original observation vector, and use the Kalman filter algorithm to calculate the Kalman gain in combination with the set state transition matrix and observation matrix to predict and update the state. These values ​​are processed to filter noise and obtain the smoothed network feature vector. The chain reaction judgment and risk assessment module is used to calculate the risk index composed of the delay accumulation deviation and the buffer data receiving speed deviation based on the obtained smooth network feature vector. If the risk index is greater than the preset threshold and the buffer data receiving speed is lower than the average level, it is determined that there is a potential interruption risk, and then a risk warning signal containing the risk index is generated. The transmission mode analysis and optimization module is used to obtain the transmission mode status of the client player's status register, which belongs to the first-play-then-push or discontinuous push, in response to the generated risk warning signal. It uses a dynamic programming algorithm to calculate the state transition cost and transmission loss, thereby optimizing the switching path between the first-play-then-push and discontinuous push, and obtaining the optimized mode switching sequence corresponding to the path with the lowest cost. The adaptation mechanism triggering and buffer adjustment module is used to extract the specific latency jitter amplitude and fluctuation frequency through the obtained mode switching sequence. If the smoothed network feature vector shows continuous fluctuation, the target buffer duration is calculated based on the latency jitter amplitude and the smoothed network feature vector, and the corresponding pull rate target is sent to the transmission control layer to maintain smooth playback. The feedback data processing and retransmission module is used to obtain reliable transmission feedback data based on the adjusted target buffer duration and pull rate target, determine whether lost packets need to be retransmitted, and if the feedback data indicates a high packet loss rate, perform a retransmission operation, calculate the round-trip time of the retransmitted packets to obtain the cumulative delay increment, and add the cumulative delay increment to the base delay to update the delay accumulation record. The chain reaction mitigation module is used to determine the reduction in latency increment caused by retransmission as the mitigation level by updating the latency accumulation record. The Kalman filter algorithm is used to recalculate the smoothing network feature vector to obtain the corrected index value. If the mitigation level is lower than the preset value, the sampling and update frequency of the Kalman filter is forcibly increased. The index correction module is used to re-execute the dynamic programming algorithm to optimize the transmission mode to ensure continuity based on the corrected index value. It calculates the playback stability index by combining the corrected index value with the actual buffer duration, and judges whether the playback smoothness has been restored based on the index. If it has not been restored, it repeatedly triggers the step of adjusting the streaming rate target until the playback stability index reaches the stable threshold.

Images