Self-adaptive low-delay motion scene live broadcast method and system

Abstract

The invention discloses a self-adaptive low-delay motion scene live broadcast method and system, and particularly relates to the technical field of scene live broadcast. Through unified mapping and exception suppression of multi-source time sequence data, a multi-scale sliding window predictor and a short-time autoregression and long-time trend sensing algorithm are combined; a more accurate bandwidth prediction result with interval confidence description is generated, characteristics such as offset cumulant, fluctuation intensity and error residence time are extracted by using a residual trajectory, threshold crossing frequency, switching amplitude, direction alternation rate and critical zone residence duration are analyzed synchronously with a parameter switching log, critical oscillation characteristics are formed, and the bandwidth prediction accuracy is improved. The risk identification is more accurate, the high-frequency oscillation risk score of the system is calculated through a normalization and time sequence risk identifier, the risk assessment result is mapped into an executable stable intervention strategy, the system state observation sequence after adjustment execution is subjected to short-time assessment, and a feedback packet is formed to write back a closed loop. And online optimization of the weight, the decision threshold and the cooling time of the bandwidth predictor is realized.

Description

Technical Field

[0001] The present invention relates to the field of scene live broadcast technology, and more specifically, to an adaptive low-latency motion scene live broadcast method and system. Background Art

[0002] In adaptive low-latency live streaming of sports scenes, the system typically relies on bandwidth prediction results to dynamically adjust key parameters such as bitrate, GOP length, and redundancy ratio to strike a balance between image quality and latency. However, in practice, network bandwidth for sports scenes often fluctuates frequently and unpredictably. When the bandwidth prediction model's update frequency or accuracy is insufficient, the predicted value lags behind the actual network conditions, resulting in accumulated bandwidth prediction errors. This error, when fed into the adaptive control process, causes the parameter adjustment logic to frequently cross set thresholds, triggering multiple rounds of bitrate, GOP, and redundancy ratio switching. Because these switching points are often concentrated near the thresholds, the system is prone to "critical oscillation"—repeatedly adjusting the bitrate up and down in a short period of time. Bandwidth prediction errors and critical oscillation effects amplify each other, ultimately causing the live broadcast link to constantly fluctuate between image quality and latency control, failing to stably maintain the desired low latency and high image quality targets, significantly degrading the viewer experience. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an adaptive low-latency motion scene live broadcast method and system to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions: The method for adaptive low-latency live broadcast of sports scenes includes the following steps: Collect multi-source operating data, generate multi-source time series data streams, perform unified time base mapping and anomaly suppression processing on the multi-source time series data streams, generate a time series consistent data set, apply a multi-scale sliding window predictor based on the time series consistent data set, combine short-term autoregressive and long-term trend perception algorithms, output bandwidth prediction values ​​and interval confidence descriptions, and form a bandwidth prediction set; The residual trajectory is calculated using the bandwidth prediction set and the actual bandwidth samples. The bias accumulation, fluctuation intensity and error residence time characteristics are extracted from the residual trajectory to generate bandwidth prediction error accumulation information. The accumulated bandwidth prediction error information is time-synchronized with the adaptive parameter switching log. The threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time are statistically analyzed to generate the adaptive parameter switching critical oscillation characteristics. The accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching are normalized and input into the timing risk identifier. The system is then given a risk score and level for high-frequency oscillations in the bitrate and latency strategies, generating a high-frequency oscillation risk assessment result. Mapping high-frequency oscillation risk assessment results into stabilization intervention strategy tokens, converting strategy tokens into executable parameter adjustment instructions, and injecting these instructions into the encoding control link and transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed; Perform a short-term evaluation on the system state observation sequence after the adjustment is executed, generate an intervention effect feedback package and write it back to the risk control closed loop.

[0005] In a preferred embodiment, unified time base mapping and anomaly suppression processing are performed on multi-source time series data streams to generate a time series consistent data set, as follows: Perform field normalization and metadata injection on multi-source original time series data streams to generate normalized time series data streams; Perform time synchronization correction and sampling reconstruction on the normalized time series data stream to generate a synchronized time series data stream; Perform anomaly detection and adaptive filtering on synchronized time series data streams to form a time-consistent data set.

[0006] In a preferred embodiment, a multi-scale sliding window predictor is applied based on the time-series consistent data set, combined with short-term autoregressive and long-term trend perception algorithms, to output bandwidth prediction values ​​and interval confidence descriptions to form a bandwidth prediction set, specifically as follows: Perform multi-scale feature extraction and exogenous fusion on the time-series consistent dataset: extract the instantaneous bandwidth gradient, instantaneous throughput volatility, and short-term packet loss rate in the short window; extract the trend slope, period baseline, and seasonal decomposition in the medium / long window; and incorporate the exogenous features to generate a multi-scale feature set. The multi-scale feature set is input into the multi-scale sliding window predictor, and the short-term autoregressive predictor, fast online regressor and long-term trend perceptron are run in parallel according to the dynamic window strategy. The prediction results of each window are adaptively weighted according to the recent residual performance to form a bandwidth prediction candidate set.

[0007] In a preferred embodiment, the bandwidth prediction set and the actual bandwidth samples are used to calculate the residual trajectory, and the bias accumulation, fluctuation intensity and error residence time characteristics are extracted from the residual trajectory to generate bandwidth prediction error accumulation information, as follows: Calculate the point-by-point residuals between the time-aligned bandwidth predictions and the actual observed bandwidth samples, and generate the initial residual trajectory; Perform multi-scale denoising and detrending on the initial residual trajectory to generate a net residual trajectory; Perform change point detection and segmented steady-state partitioning on the net residual trajectory to generate segmented residual sequences and segment boundary timestamps; In the segmented residual sequence, the intra-segment residual statistics are calculated for each segment and accumulated over time to form a bias accumulation curve, and segment-level metadata is recorded; Perform long-term offset identification and rate estimation on the bias accumulation curve to generate long-term offset identification and rate estimation; The long-term offset identification and rate estimation are combined with the segmented residual series to calculate the fluctuation intensity time series using an adaptive sliding window to generate a fluctuation intensity time series and mark it with uncertainty labels; Perform peak / cluster detection on the fluctuation intensity time series and calculate the amplitude, rise / fall slope, and duration of each peak to generate a list of error residence time features; The error retention time feature list and the bias accumulation curve are subjected to time series coupling analysis to generate an error retention and bias accumulation coupling feature set, which is then packaged into a bandwidth prediction error accumulation information package. The information package contains: bias accumulation curve, fluctuation intensity series, error retention time feature list, and error retention and bias accumulation coupling feature set.

[0008] In a preferred embodiment, the bandwidth prediction error accumulation information is time-synchronized with the adaptive parameter switching log, and the threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time are statistically analyzed to generate the adaptive parameter switching critical oscillation characteristics, as follows: Baseline-align the bandwidth prediction error accumulation information and the adaptive parameter switching log according to their original timestamps to generate a time-aligned observation sequence; Resampling and time base unification are performed on the time-aligned observation sequence, and quality marking and missing segment marking are performed on the resampled segments to generate a synchronized observation sequence with quality marking; Robust anomaly suppression and pulse retention are performed on synchronized and quality-tagged observation sequences, and confidence labels are added to the suppression / retention points to generate cleaned observation time series streams with confidence meta-information. Convert the cleaned observation time series stream with confidence information into an event stream: detect and annotate each threshold crossing, switch start / end time, and parameter values ​​before and after the switch, and generate a list of switch events with annotated before and after values; Window-pair the switching events with pre- and post-value annotations with the cleaned residual context to generate event-residual context tuples. Calculate the event-residual context tuple and add the switching amplitude index to generate an event tuple with amplitude annotation; The event tuples with amplitude annotations are directional-annotated and directional sequences are constructed: the rising / falling alternations of the continuous event sequence are identified, and the raw counts of directional alternations and the alternation rate weighted by amplitude are calculated on the sequence to generate the directional alternation rate metric; The direction alternation rate metric is combined with the event tuple to determine the critical region membership: the critical region entry / exit points are marked according to the parameter threshold and the preset hysteresis band, and continuous critical region segments are merged to calculate the critical region retention time and generate a critical region event list; The adaptive parameter switching critical oscillation feature set is summarized, including: threshold crossing frequency, normalized average switching amplitude, amplitude-weighted direction alternation rate, and average critical zone residence time; and a quality label and time window identifier are attached to each item to form a critical oscillation feature package that can be directly input into the downstream risk assessor.

[0009] In a preferred embodiment, the bandwidth prediction error accumulation information and the adaptive parameter switching critical oscillation characteristics are normalized and input into a timing risk identifier, which outputs a risk score and level of the system being in a high-frequency oscillation state of bit rate and delay strategy, thereby generating a high-frequency oscillation risk assessment result. The risk score of the system being in a high-frequency oscillation state of the bit rate and delay strategy is obtained by weighted summing the bandwidth prediction error accumulation information and the adaptive parameter switching critical oscillation characteristics.

[0010] In a preferred embodiment, the adaptive low-latency sports scene live broadcast system includes a multi-source timing prediction processing module, a bandwidth prediction error analysis module, an adaptive parameter switching critical oscillation analysis module, a high-frequency oscillation risk assessment module, a stabilization intervention strategy execution module, and an intervention effect feedback and closed-loop update module; A multi-source time series prediction processing module is used to collect multi-source operation data, generate multi-source time series data streams, perform unified time base mapping and anomaly suppression processing on the multi-source time series data streams, generate a time series consistent data set, apply a multi-scale sliding window predictor based on the time series consistent data set, combine short-term autoregressive and long-term trend perception algorithms, output bandwidth prediction values ​​and interval confidence descriptions, and form a bandwidth prediction set; The bandwidth prediction error analysis module is used to calculate the residual trajectory using the bandwidth prediction set and the actual bandwidth samples, and extract the bias accumulation, fluctuation intensity and error residence time characteristics from the residual trajectory to generate bandwidth prediction error accumulation information; The adaptive parameter switching critical oscillation analysis module is used to synchronize the accumulated bandwidth prediction error information with the adaptive parameter switching log, calculate the threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time, and generate the adaptive parameter switching critical oscillation characteristics; The high-frequency oscillation risk assessment module is used to normalize the accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching, and input them into the timing risk identifier. It outputs the risk score and level of the system in the high-frequency oscillation state of bit rate and delay strategy, and generates the high-frequency oscillation risk assessment result; A stabilization intervention strategy execution module is used to map high-frequency oscillation risk assessment results into stabilization intervention strategy tokens, convert strategy tokens into executable parameter adjustment instructions, and inject the instructions into the encoding control link and transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed; The intervention effect feedback and closed-loop update module is used to perform short-term evaluation of the system state observation sequence after the adjustment is executed, generate the intervention effect feedback package and write it back to the risk control closed loop.

[0011] The technical effects and advantages of the present invention are as follows: 1. The present invention integrates bandwidth prediction error accumulation analysis with adaptive parameter switching critical oscillation identification to achieve dynamic regulation of key parameters such as bit rate, GOP length, and redundancy rate in live broadcast links of sports scenes. It can effectively suppress the occurrence of high-frequency oscillation states, allowing the system to maintain a balance between image quality and latency under conditions of severe network bandwidth fluctuations. Through unified mapping and anomaly suppression of multi-source time series data, combined with a multi-scale sliding window predictor and short-term autoregressive and long-term trend perception algorithms, it generates more accurate bandwidth prediction results with interval confidence descriptions, thereby significantly reducing prediction error accumulation. At the same time, the residual trajectory is used to extract features such as bias accumulation, fluctuation intensity, and error retention time. The threshold crossing frequency, switching amplitude, direction alternation rate and critical zone retention time are analyzed synchronously with the parameter switching log to form critical oscillation characteristics, making risk identification more accurate. The system high-frequency oscillation risk score is calculated through normalization and timing risk identifier, and the risk assessment result is mapped into an executable stabilization intervention strategy. It is then injected into the coding and transmission control link through delay-sensitive smooth scheduling, which can smoothly transition parameter changes during the adjustment execution process and avoid image quality jitter or delay surges caused by sudden changes. Finally, the system state observation sequence after the adjustment is short-term evaluated and a feedback packet write-back closed loop is formed to achieve online optimization of the bandwidth predictor weight, decision threshold and cooling time.

[0012] 2. The present invention can significantly improve the stability and reliability of the live broadcast system in sports scenarios, reduce the risk of high-frequency fluctuations in image quality and delay, enable viewers to obtain a continuous high-quality, low-latency viewing experience, and at the same time enhance the system's adaptability to complex network environments and self-regulation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flow chart of the method of Example 1 of the present invention; Figure 2 This is a flow chart of the system of Example 2 of the present invention. DETAILED DESCRIPTION

[0014] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0015] Example 1: Figure 1 The present invention provides an adaptive low-latency live broadcast method for sports scenes, comprising the following steps: Collect multi-source operating data, generate multi-source time series data streams, perform unified time base mapping and anomaly suppression processing on the multi-source time series data streams, generate a time series consistent data set, apply a multi-scale sliding window predictor based on the time series consistent data set, combine short-term autoregressive and long-term trend perception algorithms, output bandwidth prediction values ​​and interval confidence descriptions, and form a bandwidth prediction set; The residual trajectory is calculated using the bandwidth prediction set and the actual bandwidth samples. The bias accumulation, fluctuation intensity and error residence time characteristics are extracted from the residual trajectory to generate bandwidth prediction error accumulation information. The accumulated bandwidth prediction error information is time-synchronized with the adaptive parameter switching log. The threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time are statistically analyzed to generate the adaptive parameter switching critical oscillation characteristics. The accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching are normalized and input into the timing risk identifier. The system is then given a risk score and level for high-frequency oscillations in the bitrate and latency strategies, generating a high-frequency oscillation risk assessment result. Mapping high-frequency oscillation risk assessment results into stabilization intervention strategy tokens, converting strategy tokens into executable parameter adjustment instructions, and injecting these instructions into the encoding control link and transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed; Conduct a short-term evaluation of the system state observation sequence after the adjustment is executed, generate an intervention effect feedback package and write it back to the risk control closed loop; Collect multi-source operation data (including encoder output bit rate and frame motion vector distribution, transmission link instantaneous bandwidth samples, round-trip delay / one-way delay, packet loss / retransmission count, receiver buffer occupancy and decoding delay, etc.) to generate multi-source time series data streams.

[0016] Perform unified time base mapping and anomaly suppression on multi-source time series data streams to generate a time series consistent data set, as follows: Perform field normalization and metadata injection (additional sampling precision, timestamp source identifier, path label, device identifier, and sample quality tag) on ​​multi-source original time series data streams to generate normalized time series data streams; Perform time synchronization correction and sampling reconstruction on the normalized time series data stream (perform clock drift correction, time base difference alignment, interpolation and downsampling when necessary, mark missing segments and record interpolation confidence) to generate a synchronized time series data stream; Perform anomaly detection and adaptive filtering on synchronized time series data streams (implementing isolated point identification, short-time jump suppression, burst retention strategy and local smoothing, increasing the sampling rate based on event triggers to retain key burst signals, and attaching quality confidence annotations to each data segment) to form a time-consistent data set; Based on the time-series consistent dataset, a multi-scale sliding window predictor is applied, combined with short-term autoregressive and long-term trend perception algorithms, to output bandwidth prediction values ​​and interval confidence descriptions to form a bandwidth prediction set, as follows: Multi-scale feature extraction and exogenous fusion are performed on the time-consistent dataset: instantaneous bandwidth gradient, instantaneous throughput volatility, and short-term packet loss rate are extracted in the short window; trend slope, periodic baseline, and seasonal decomposition are extracted in the medium / long window; exogenous features such as motion complexity and view switching events are incorporated to generate a multi-scale feature set; The multi-scale feature set is input into the multi-scale sliding window predictor. The short-term autoregressive predictor, fast online regressor, and long-term trend sensor are run in parallel according to the dynamic window strategy (sudden change / shift detection is performed in parallel to identify state mutations). The prediction results of each window are adaptively weighted according to the recent residual performance to form a bandwidth prediction candidate set. Uncertainty estimation and interval confidence description are performed on the bandwidth prediction candidate set (quantile intervals are generated using methods such as model ensemble distribution, bootstrap sampling, or approximate Bayesian posterior estimation, and online calibration is performed using historical coverage). This generates a bandwidth prediction set and the corresponding interval confidence description and confidence label. The bandwidth prediction set is verified for confidence applicability and annotated with meta-information (based on recent residual statistics to verify confidence interval coverage, trigger conservative predictors in high uncertainty windows and mark them, and record predictor priorities and applicable time windows). The final bandwidth prediction set and confidence meta-information for downstream use are output, and the predictions and actual observations are recorded in parallel for subsequent online reweighting and offline retraining.

[0017] The residual trajectory is calculated using the bandwidth prediction set and the actual bandwidth samples. The bias accumulation, fluctuation intensity, and error residence time characteristics are extracted from the residual trajectory to generate bandwidth prediction error accumulation information, as follows: Calculate the point-by-point residual of the time-aligned bandwidth prediction and the actual observed bandwidth sample (at time point t, the observed bandwidth value b_obs(t) is subtracted from the bandwidth prediction value b_pred(t), and the point-by-point residual r(t) is equal to the observed value minus the predicted value), and generate the initial residual trajectory; Perform multi-scale denoising and detrending processing (short-window noise suppression, medium- and long-term baseline separation, and retention of sudden pulses) on the initial residual trajectory to generate a net residual trajectory; Perform change point detection and segmented steady-state partitioning on the net residual trajectory (using online change point detection and minimum description length criterion) to generate segmented residual sequences and segment boundary timestamps; In the segmented residual sequence, the intra-segment residual statistics are calculated for each segment and accumulated over time to form a bias accumulation curve, and segment-level metadata (including intra-segment mean accumulation, inter-segment offset direction, and accumulation duration) are recorded. Perform long-term drift identification and rate estimation on the bias accumulation curve (determine long-term drift based on inter-segment trend consistency and cumulative slope), and generate long-term drift identification and rate estimation; The long-term offset identification and rate estimation of the offset accumulation curve includes: The bias accumulation curve is preliminarily smoothed according to the preset minimum segment length to suppress high-frequency noise and generate a smooth bias curve; Perform seasonal and cyclical decomposition on the smoothed bias curve (extract baseline trend, cyclical component, and residual component) to generate baseline trend series and cyclical component annotations; Re-segment the baseline trend sequence according to the previous segmentation boundaries to generate a segmented baseline sequence and the start and end timestamps of each segment; Calculate a robust slope estimate for each segment of the segmented baseline series (using anomaly-resistant regression methods such as Theil–Sen or robust least squares), and generate a segment-level slope set and segment-level slope confidence interval; Perform a directional consistency check on the segment-level slope set (determine whether the slope signs of most segments are consistent) and calculate the slope sign ratio and the transition rate of the slope signs of adjacent segments to generate an inter-segment trend consistency measure (the trend consistency measure is obtained by the weighted sum of the slope sign ratio and the transition rate of the slope signs of adjacent segments). The segment-level slope set is synthesized by cumulative slope according to time weight (giving higher weight to recent segments or weighting according to segment length / energy) to generate a cumulative slope curve; Perform significance tests on the cumulative slope curve (using non-parametric trend tests or bootstrap sampling methods to assess the significance of the cumulative slope) to generate long-term shift significance indicators and confidence intervals; The long-term shift significance index and the inter-segment trend consistency measure are combined to determine the long-term shift flag: when the significance passes and the trend consistency exceeds the threshold, the long-term shift flag is set to "exist", otherwise it is set to "not detected"; the long-term shift flag is generated; When the cumulative slope curve is confirmed to be "present", a rate estimate is performed (the rate estimate is obtained by weighted average of the segment slopes); The long-term offset marker and rate estimate are combined with the segmented residual series to calculate the fluctuation intensity time series using an adaptive sliding window (the window length is adaptively adjusted according to historical fluctuations, and the indicator uses robust variance / interquartile range or energy spectrum estimation). This generates a fluctuation intensity time series and adds an uncertainty label. Perform peak / cluster detection on the fluctuation intensity time series and count the amplitude, rise / fall slope, and duration of each peak to generate a list of error residence time features (recording the peak start and end, peak energy, and duration window); The error retention time feature list and the bias accumulation curve are subjected to time series coupling analysis (quantified using the Pearson correlation coefficient) to generate an error retention and bias accumulation coupling feature set (including the Pearson correlation coefficient and relative time series). The feature set is then packaged into a bandwidth prediction error accumulation information package, which includes: bias accumulation curve, fluctuation intensity series, error retention time feature list, and error retention and bias accumulation coupling feature set.

[0018] The accumulated bandwidth prediction error information is time-synchronized with the adaptive parameter switching log. The threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time are statistically analyzed to generate the adaptive parameter switching critical oscillation characteristics, as follows: Baseline-align the bandwidth prediction error accumulation information and the adaptive parameter switching log according to their original timestamps to generate a time-aligned observation sequence; Resampling and time base unification are performed on the time-aligned observation sequences (using the required minimum resolution or dynamic time warping calibration), and quality marking and missing segment marking are performed on the resampled segments to generate synchronized observation sequences with quality markings. Robust anomaly suppression and pulse retention (using median filtering combined with event retention strategy) are performed on synchronized, quality-tagged observation sequences, and confidence labels are added to the suppression / retention points to generate cleaned observation time series streams with confidence meta-information. Convert the cleaned observation time series stream with confidence information into an event stream: detect and annotate each threshold crossing (including the up and down crossing direction), the switching start / end time, and the parameter values ​​before and after the switching, and generate a switching event list with the before and after value annotations; The switching events with pre- and post-value annotations are paired with the cleaned residual context in a windowed manner (the adaptive window length is adjusted based on the local fluctuation intensity, and the window includes the short-term baseline before the switch and the short-term response after the switch) to generate an event-residual context tuple. It should be noted that the cleaned residual context refers to the local time period data retained from the bandwidth prediction residual trajectory after outlier suppression, noise removal, and trend separation. This time period is centered on a specific event (such as adaptive parameter switching) and includes the residual values, trend direction, fluctuation intensity, etc. before and after the event. It not only contains the numerical sequence of residuals, but also retains the quality labels related to the time period (such as "trusted after suppression" / "partially missing") and processing meta-information (denoising method, filter window length, etc.). Its purpose is to provide a clean and contextually meaningful reference window when analyzing the relationship between switching events and residual changes, preventing original noise from interfering with judgment.

[0019] Calculate the event-residual context tuple and append the switching amplitude indicator (absolute amplitude and normalized amplitude relative to the baseline) to generate an event tuple with amplitude annotation; The event tuples with amplitude annotations are directional-annotated and directional sequences are constructed: the rising / falling alternations of the continuous event sequence are identified, and the raw counts of directional alternations and the alternation rate weighted by amplitude are calculated on the sequence to generate the directional alternation rate metric; The direction alternation rate metric is combined with the event tuple to determine the critical region membership: the critical region entry / exit points are marked according to the parameter threshold and the preset hysteresis band, and continuous critical region segments are merged to calculate the critical region retention time and generate a critical region event list; It should be noted that the hysteresis band refers to a dual-threshold interval or buffer zone used to prevent parameters from switching back and forth repeatedly due to slight fluctuations in adaptive switching. It includes the threshold for entering the critical zone (triggering switching) and the threshold for exiting the critical zone (releasing switching). There is a certain interval (bandwidth or delay value difference) between these two thresholds, forming a buffer area in the "bandwidth-delay space". Within this interval, even if the monitoring value fluctuates back and forth, switching will not be triggered immediately, thereby reducing the probability of "jitter" or "critical oscillation". The hysteresis band is identified to determine the system's residence time in the critical zone, thereby assessing whether it is in a high-risk oscillation state.

[0020] The direction alternation rate metric is combined with the event tuple to determine the critical section membership and generate a critical section event list, as follows: Align the direction alternation rate metric with the event tuples one by one according to the time index, and embed the corresponding alternation rate value in each event tuple to generate an event sequence with direction alternation rate labels; The event sequence with direction alternation rate labels is scanned in event time order, and the event parameter values ​​and their change directions are extracted during the scanning process. They are compared point by point with the preset parameter threshold interval and hysteresis band to generate a "critical state mark" (entry, exit, or maintenance) for each event; It should be noted that the event parameter value refers to the specific value of the core operating parameter adjusted in the system before and after a certain adaptive parameter switching event occurs.

[0021] For example: Coding parameters: video encoding bitrate (e.g., 3.2 Mbps to 2.5 Mbps); GOP length (e.g., 60 frames to 30 frames); encoding quantization parameter (QP) value (e.g., 28 to 32); encoding resolution (e.g., 1920×1080 to 1280×720); Transmission-related parameters: target sender throughput (e.g., 5 Mbps to 4 Mbps); FEC redundancy ratio (e.g., 10% to 15%); target adaptive buffer length (e.g., 300 ms to 250 ms); Player-side related parameters (if bidirectional adaptation is available): playback buffer target delay (e.g., 200 ms changed to 180 ms); player-side decoding frame rate limit (e.g., 60 fps changed to 48 fps).

[0022] Track the state transition of each event's critical state marker (from entering change to maintaining change to exiting), and use the upper and lower boundaries of the hysteresis band to suppress frequent state jitter caused by measurement noise to generate a noise-suppressed critical state time series; Merge continuous segments of the critical state time series after noise suppression (merging rule: adjacent state segments with intervals less than the minimum interval threshold are merged into the same segment) to generate a list of candidate critical region segments; Cross-screen the list of candidate segments of the critical region with the direction alternation pattern of the event sequence (segments with a direction alternation rate higher than the set percentile threshold are retained) to generate a filtered set of valid segments of the critical region; The retention time of each segment of the critical region valid segment set is calculated (segment end time minus segment start time), and the following statistics are added: maximum retention time, average retention time, and retention time variance to generate the critical region segment retention feature set; Jointly annotate the critical zone segment retention feature set with the alternation rate and switching amplitude distribution within the segment (generating a multi-dimensional segment profile, including retention intensity, direction alternation activity, and amplitude fluctuation intensity), forming a structured critical zone event list as a direct input for downstream risk identification and strategy generation; The adaptive parameter switching critical oscillation feature set is summarized, including: threshold crossing frequency, normalized average switching amplitude, amplitude-weighted direction alternation rate, and average critical zone residence time; and a quality label and time window identifier are attached to each item to form a critical oscillation feature package that can be directly input into the downstream risk assessor.

[0023] The accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching are normalized and input into the timing risk identifier. The system is then given a risk score and level for high-frequency oscillations in the bitrate and latency strategies, generating a high-frequency oscillation risk assessment result. The risk score of the system being in a high-frequency oscillation state of the bit rate and delay strategy is obtained by weighted summing the bandwidth prediction error accumulation information and the adaptive parameter switching critical oscillation characteristics.

[0024] Compare the risk score of the system in the state of high-frequency fluctuation of bit rate and delay strategy with the preset scoring stage threshold to classify the high-frequency fluctuation risk; The grading results are combined with the risk score to form the high-frequency oscillation risk assessment results.

[0025] The high-frequency oscillation risk assessment results are mapped to stabilization intervention strategy tokens, which are then converted into executable parameter adjustment instructions. These instructions are then injected into the encoding control link and the transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed, as follows: Map the high-frequency shock risk assessment results into a set of policy tokens. Each token contains (risk level, list of recommended measures, set of impact parameters, upper limit of action duration, and safety margin), and generate a set of policy tokens. Prioritize the policy token set according to risk level, current system SLA, and business priority. Add parameter constraints (maximum allowable amplitude, minimum residence time, mutual exclusion rules) to each token based on the security boundary library and dependency graph to generate a constrained policy token queue. The constrained policy token queue is mapped to a predefined parameter template library to generate a set of parameter adjustment drafts. The drafts include recommended values ​​and execution windows for the encoder (target bitrate cap, QP change, GOP / keyframe interval, keyframe priority, and redundancy ratio) and the transmitter (packet scheduling priority, fragmentation / retransmission strategy, FEC ratio, path selection recommendation, and distribution rate cap). Perform constraint checking (boundary violation detection, mutual exclusion conflict resolution, and fallback path confirmation) on the parameter adjustment draft set. When conflicts arise, resolve them based on priority and dependency relationships (for example, replacing latency-sensitive measures with image quality degradation measures), generating a constrained set of operation instructions with a rollback strategy. The operation instruction set is decomposed into time segments according to instruction priority, impact scope, and minimum interruption principle. Each instruction is assigned an execution start and end window, a maximum execution slope (maximum parameter change rate), and a command frequency limit (token bucket style) to form a scheduling schedule to be injected. A delay-sensitive smoothing scheduling algorithm is applied to the schedule. This algorithm uses (priority, deadline, system delay budget, maximum impact budget, and minimum residence time) as scheduling constraints. It uses a piecewise linear / smooth sigmoid slope or exponential decay function to generate a progressive parameter curve, and reserves a short preemption window before high-confidence / high-priority instructions. The algorithm outputs a progressive execution plan and packages it into a command package with a version number. Convert each progressive execution plan into commands recognizable by the encoding control (e.g., setting the instantaneous bitrate cap, submitting QP adjustments, keyframe insertion commands, adjusting the FEC ratio, setting the transmission queue priority), and use atomic or phased submission strategies (including pre-submit-confirm-validate three-stage or two-stage submission) to issue command packets, generating execution confirmation and initial response indication; When the confirmation and initial response indication trigger operation is executed, the system is sampled at high frequency according to the sampler to collect and record key observation quantities (real-time bit rate output, actual GOP / key frame time, end-to-end delay distribution, packet loss / replay count, FEC effectiveness, buffer occupancy, switching event count, etc.), and form a system status observation sequence after adjustment execution in time series.

[0026] Perform a short-term evaluation of the system state observation sequence after the adjustment is executed, generate an intervention effect feedback package and write it back to the risk control closed loop, as follows: Perform a short-term evaluation of the system state observation sequence after the adjustment (comparing expected and actual latency / image quality / stability indicators). If the observation sequence triggers a rollback condition (such as latency exceeding the budget, jitter amplification, or critical service SLA violation), roll back to the previous confirmable state according to the rollback policy in the instruction package and generate a rollback record. Otherwise, the confirmation takes effect and generates the final execution state. The final execution status and observation sequence generate an intervention effect feedback package and write it back to the risk control closed loop: the final execution status, observation sequence summary, instruction execution log and rollback record are packaged into an intervention effect feedback package, and written back to the prediction / risk assessment and strategy priority maintenance process according to the time window and quality label. The intervention effect feedback package and the corresponding policy token, operation instruction set and execution log are versioned and archived and marked with security audits to form a traceable policy execution history record.

[0027] The present invention integrates bandwidth prediction error accumulation analysis and adaptive parameter switching critical oscillation identification to achieve dynamic regulation of key parameters such as bit rate, GOP length and redundancy rate in live broadcast links of sports scenes. It can effectively suppress the occurrence of high-frequency oscillation states, so that the system can still maintain a balance between image quality and delay under conditions of severe network bandwidth fluctuations. Through unified mapping and anomaly suppression of multi-source time series data, combined with a multi-scale sliding window predictor and short-term autoregressive and long-term trend perception algorithms, a more accurate bandwidth prediction result with interval confidence description is generated, thereby significantly reducing the accumulation of prediction errors. At the same time, the residual trajectory is used to extract features such as bias accumulation, fluctuation intensity and error retention time. The threshold crossing frequency, switching amplitude, direction alternation rate and critical zone retention time are analyzed synchronously with the parameter switching log to form critical oscillation characteristics, making risk identification more accurate. The system high-frequency oscillation risk score is calculated through normalization and timing risk identifier, and the risk assessment result is mapped into an executable stabilization intervention strategy. It is then injected into the encoding and transmission control link through delay-sensitive smooth scheduling, which can smoothly transition parameter changes during the adjustment execution process and avoid image quality jitter or delay spikes caused by sudden changes. Finally, the system state observation sequence after the adjustment is short-term evaluated and a feedback packet write-back closed loop is formed to achieve online optimization of the bandwidth predictor weight, decision threshold and cooling time.

[0028] The present invention can significantly improve the stability and reliability of the live broadcast system in sports scenarios, reduce the risk of high-frequency fluctuations in image quality and delay, enable viewers to obtain a continuously high-quality, low-latency viewing experience, and at the same time enhance the system's adaptability to complex network environments and self-regulation efficiency.

[0029] Example 2: This example is an introduction to the adaptive low-latency live broadcast system for sports scenes. Figure 2 As shown, it includes a multi-source time series prediction processing module, a bandwidth prediction error analysis module, an adaptive parameter switching critical oscillation analysis module, a high-frequency oscillation risk assessment module, a stabilization intervention strategy execution module, and an intervention effect feedback and closed-loop update module; A multi-source time series prediction processing module is used to collect multi-source operation data, generate multi-source time series data streams, perform unified time base mapping and anomaly suppression processing on the multi-source time series data streams, generate a time series consistent data set, apply a multi-scale sliding window predictor based on the time series consistent data set, combine short-term autoregressive and long-term trend perception algorithms, output bandwidth prediction values ​​and interval confidence descriptions, and form a bandwidth prediction set; The bandwidth prediction error analysis module is used to calculate the residual trajectory using the bandwidth prediction set and the actual bandwidth samples, and extract the bias accumulation, fluctuation intensity and error residence time characteristics from the residual trajectory to generate bandwidth prediction error accumulation information; The adaptive parameter switching critical oscillation analysis module is used to synchronize the accumulated bandwidth prediction error information with the adaptive parameter switching log, calculate the threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time, and generate the adaptive parameter switching critical oscillation characteristics; The high-frequency oscillation risk assessment module is used to normalize the accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching, and input them into the timing risk identifier. It outputs the risk score and level of the system in the high-frequency oscillation state of bit rate and delay strategy, and generates the high-frequency oscillation risk assessment result; A stabilization intervention strategy execution module is used to map high-frequency oscillation risk assessment results into stabilization intervention strategy tokens, convert strategy tokens into executable parameter adjustment instructions, and inject the instructions into the encoding control link and transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed; The intervention effect feedback and closed-loop update module is used to perform a short-term evaluation of the system state observation sequence after the adjustment is executed, generate an intervention effect feedback package, and write it back to the risk control closed loop; The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0030] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0031] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0032] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process of the system and method described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0033] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways.

[0034] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An adaptive low-latency live broadcast method for sports scenes, characterized by: The steps include: Collect multi-source operating data, generate multi-source time series data streams, perform unified time base mapping and anomaly suppression processing on the multi-source time series data streams, generate a time series consistent data set, apply a multi-scale sliding window predictor based on the time series consistent data set, combine short-term autoregressive and long-term trend perception algorithms, output bandwidth prediction values ​​and interval confidence descriptions, and form a bandwidth prediction set; The residual trajectory is calculated using the bandwidth prediction set and the actual bandwidth samples. The bias accumulation, fluctuation intensity and error residence time characteristics are extracted from the residual trajectory to generate bandwidth prediction error accumulation information. The accumulated bandwidth prediction error information is time-synchronized with the adaptive parameter switching log. The threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time are statistically analyzed to generate the adaptive parameter switching critical oscillation characteristics. The accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching are normalized and input into the timing risk identifier. The system is then given a risk score and level for high-frequency oscillations in the bitrate and latency strategies, generating a high-frequency oscillation risk assessment result. Mapping high-frequency oscillation risk assessment results into stabilization intervention strategy tokens, converting strategy tokens into executable parameter adjustment instructions, and injecting these instructions into the encoding control link and transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed; Perform a short-term evaluation on the system state observation sequence after the adjustment is executed, generate an intervention effect feedback package and write it back to the risk control closed loop.

2. The adaptive low-latency live broadcast method for sports scenes according to claim 1, characterized in that: Perform unified time base mapping and anomaly suppression on multi-source time series data streams to generate a time series consistent data set, as follows: Perform field normalization and metadata injection on multi-source original time series data streams to generate normalized time series data streams; Perform time synchronization correction and sampling reconstruction on the normalized time series data stream to generate a synchronized time series data stream; Perform anomaly detection and adaptive filtering on synchronized time series data streams to form a time-consistent data set.

3. The adaptive low-latency live broadcast method for sports scenes according to claim 2, characterized in that: Based on the time-series consistent dataset, a multi-scale sliding window predictor is applied, combined with short-term autoregressive and long-term trend perception algorithms, to output bandwidth prediction values ​​and interval confidence descriptions to form a bandwidth prediction set, as follows: Perform multi-scale feature extraction and exogenous fusion on the time-series consistent dataset: extract instantaneous bandwidth gradient, instantaneous throughput volatility, and short-term packet loss rate in short windows; extract trend slope, period baseline, and seasonal decomposition in medium / long windows; and incorporate exogenous features; Generate multi-scale feature sets; The multi-scale feature set is input into the multi-scale sliding window predictor, and the short-term autoregressive predictor, fast online regressor and long-term trend perceptron are run in parallel according to the dynamic window strategy. The prediction results of each window are adaptively weighted according to the recent residual performance to form a bandwidth prediction candidate set.

4. The adaptive low-latency live broadcast method for sports scenes according to claim 3, characterized in that: The residual trajectory is calculated using the bandwidth prediction set and the actual bandwidth samples. The bias accumulation, fluctuation intensity, and error residence time characteristics are extracted from the residual trajectory to generate bandwidth prediction error accumulation information, as follows: Calculate the point-by-point residuals between the time-aligned bandwidth predictions and the actual observed bandwidth samples, and generate the initial residual trajectory; Perform multi-scale denoising and detrending on the initial residual trajectory to generate a net residual trajectory; Perform change point detection and segmented steady-state partitioning on the net residual trajectory to generate segmented residual sequences and segment boundary timestamps; In the segmented residual sequence, the intra-segment residual statistics are calculated for each segment and accumulated over time to form a bias accumulation curve, and segment-level metadata is recorded; Perform long-term offset identification and rate estimation on the bias accumulation curve to generate long-term offset identification and rate estimation; The long-term offset identification and rate estimation are combined with the segmented residual series to calculate the fluctuation intensity time series using an adaptive sliding window to generate a fluctuation intensity time series and mark it with uncertainty labels; Perform peak / cluster detection on the fluctuation intensity time series and calculate the amplitude, rise / fall slope, and duration of each peak to generate a list of error residence time features; The error retention time feature list and the bias accumulation curve are subjected to time series coupling analysis to generate an error retention and bias accumulation coupling feature set, which is then packaged into a bandwidth prediction error accumulation information package. The information package contains: bias accumulation curve, fluctuation intensity series, error retention time feature list, and error retention and bias accumulation coupling feature set.

5. The adaptive low-latency live broadcast method for sports scenes according to claim 4, characterized in that: The accumulated bandwidth prediction error information is time-synchronized with the adaptive parameter switching log. The threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time are statistically analyzed to generate the adaptive parameter switching critical oscillation characteristics, as follows: Baseline-align the bandwidth prediction error accumulation information and the adaptive parameter switching log according to their original timestamps to generate a time-aligned observation sequence; Resampling and time base unification are performed on the time-aligned observation sequence, and quality marking and missing segment marking are performed on the resampled segments to generate a synchronized observation sequence with quality marking; Robust anomaly suppression and pulse retention are performed on synchronized and quality-tagged observation sequences, and confidence labels are added to the suppression / retention points to generate cleaned observation time series streams with confidence meta-information. Convert the cleaned observation time series stream with confidence information into an event stream: detect and annotate each threshold crossing, switch start / end time, and parameter values ​​before and after the switch, and generate a list of switch events with annotated before and after values; Window-pair the switching events with pre- and post-value annotations with the cleaned residual context to generate event-residual context tuples. Calculate the event-residual context tuple and add the switching amplitude index to generate an event tuple with amplitude annotation; The event tuples with amplitude annotations are directional-annotated and directional sequences are constructed: the rising / falling alternations of the continuous event sequence are identified, and the raw counts of directional alternations and the alternation rate weighted by amplitude are calculated on the sequence to generate the directional alternation rate metric; The direction alternation rate metric is combined with the event tuple to determine the critical region membership: the critical region entry / exit points are marked according to the parameter threshold and the preset hysteresis band, and continuous critical region segments are merged to calculate the critical region retention time and generate a critical region event list; Summarize the adaptive parameter switching critical oscillation feature set, including: threshold crossing frequency, normalized average switching amplitude, amplitude-weighted direction alternation rate, and average critical zone residence time; Each item is accompanied by a quality label and time window identifier to form a critical oscillation feature package that can be directly input into downstream risk assessors.

6. The adaptive low-latency live broadcast method for sports scenes according to claim 5, characterized in that: The accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching are normalized and input into the timing risk identifier. The system is then given a risk score and level for high-frequency oscillations in the bitrate and latency strategies, generating a high-frequency oscillation risk assessment result. The risk score of the system being in a high-frequency oscillation state of the bit rate and delay strategy is obtained by weighted summing the bandwidth prediction error accumulation information and the adaptive parameter switching critical oscillation characteristics.

7. An adaptive low-latency sports scene live broadcast system, configured to implement the adaptive low-latency sports scene live broadcast method according to any one of claims 1 to 6, characterized in that: It includes a multi-source time series prediction processing module, a bandwidth prediction error analysis module, an adaptive parameter switching critical oscillation analysis module, a high-frequency oscillation risk assessment module, a stabilization intervention strategy execution module, and an intervention effect feedback and closed-loop update module. A multi-source time series prediction processing module is used to collect multi-source operation data, generate multi-source time series data streams, perform unified time base mapping and anomaly suppression processing on the multi-source time series data streams, generate a time series consistent data set, apply a multi-scale sliding window predictor based on the time series consistent data set, combine short-term autoregressive and long-term trend perception algorithms, output bandwidth prediction values ​​and interval confidence descriptions, and form a bandwidth prediction set; The bandwidth prediction error analysis module is used to calculate the residual trajectory using the bandwidth prediction set and the actual bandwidth samples, and extract the bias accumulation, fluctuation intensity and error residence time characteristics from the residual trajectory to generate bandwidth prediction error accumulation information; The adaptive parameter switching critical oscillation analysis module is used to synchronize the accumulated bandwidth prediction error information with the adaptive parameter switching log, calculate the threshold crossing frequency, switching amplitude, direction alternation rate, and critical zone retention time, and generate the adaptive parameter switching critical oscillation characteristics; The high-frequency oscillation risk assessment module is used to normalize the accumulated bandwidth prediction error information and the critical oscillation characteristics of adaptive parameter switching, and input them into the timing risk identifier. It outputs the risk score and level of the system in the high-frequency oscillation state of bit rate and delay strategy, and generates the high-frequency oscillation risk assessment result; A stabilization intervention strategy execution module is used to map high-frequency oscillation risk assessment results into stabilization intervention strategy tokens, convert strategy tokens into executable parameter adjustment instructions, and inject the instructions into the encoding control link and transmission scheduling link using a delay-sensitive smoothing scheduling algorithm to generate a system state observation sequence after the adjustment is executed; The intervention effect feedback and closed-loop update module is used to perform short-term evaluation of the system state observation sequence after the adjustment is executed, generate the intervention effect feedback package and write it back to the risk control closed loop.

Images