An edge computing-based video conference low-delay streaming transmission optimization method

Abstract

This invention relates to the field of information transmission technology, and more particularly to a low-latency streaming media transmission optimization method for video conferencing based on edge computing. First, based on the collected raw latency values, a predicted latency value for the path is obtained. Then, based on the predicted latency value, a dynamic smoothing mechanism based on path structure awareness is used to smooth the path, resulting in a smoothed predicted latency value. Next, based on the smoothed predicted latency value, a composite cost function is used to obtain the comprehensive cost of the path. Based on the comprehensive cost, the optimal transmission path is determined. Based on the optimal transmission path, a comprehensive feedback video quality score calculation formula is introduced to calculate the video quality score. Finally, based on the video quality score, video parameters are adjusted. This method solves the technical problems of uncontrollable end-to-end latency, unstable path selection, and lag in playback quality adjustment during streaming media transmission.

Description

Technical Field

[0001] This invention relates to the field of information transmission technology, and in particular to a method for optimizing low-latency streaming media transmission in video conferencing based on edge computing. Background Technology

[0002] With the rapid development of remote collaborative work, online education, and telemedicine, video conferencing, as a core communication method, is increasingly used, placing higher demands on its real-time performance, stability, and service quality. The core of a video conferencing system lies in ensuring low-latency, high-quality transmission of audio and video data between multiple terminals and maintaining a smooth interactive experience in dynamically changing network environments. However, traditional methods are mostly based on centralized architectures or static path scheduling mechanisms, relying on preset strategies or real-time measurement feedback for path selection, making it difficult to anticipate and intelligently respond to changes in network conditions. In actual operation, especially in public networks, cross-regional scenarios, or wireless networks, path transmission delays and bandwidth fluctuations are frequent, and sudden packet loss, latency jitter, and path instability become major factors affecting user experience. Existing streaming media scheduling mechanisms mainly adjust links through methods such as RTCP feedback, congestion control algorithms (e.g., Google Congestion Control), or application-layer multi-path selection. These methods generally adopt passive scheduling strategies, only making reactive adjustments after significant network anomalies occur, lacking predictive capabilities based on historical trends. Meanwhile, existing methods are prone to making incorrect judgments when network conditions fluctuate briefly, triggering frequent path switching and drastic parameter adjustments, which further exacerbates jitter, stuttering, and reconnection at the playback end. Furthermore, existing methods typically ignore client playback status, such as buffer occupancy and frame rate fluctuations, affecting the user's subjective experience and failing to achieve truly experience-driven transmission optimization. With the rapid deployment of new network architectures such as edge computing, multi-access edge computing, and SD-WAN, leveraging the proximity processing capabilities of edge nodes to achieve path-structure-oriented latency prediction, path stability modeling, and dynamic optimization driven by playback quality feedback has become a key technical direction for improving the stability and service quality of video conferencing systems. Therefore, there is an urgent need to propose a low-latency streaming media transmission optimization method for video conferencing based on edge computing. Summary of the Invention

[0003] This invention provides a low-latency streaming media transmission optimization method for video conferencing based on edge computing, in order to solve the problems of uncontrollable end-to-end latency, unstable path selection, and lagging playback quality adjustment that exist in existing methods during streaming media transmission in variable and unstable network environments.

[0004] The present invention provides a low-latency streaming media transmission optimization method for video conferencing based on edge computing, comprising the following steps: S1. Based on the collected raw delay value, the predicted delay value of the path is obtained; based on the predicted delay value of the path, a dynamic smoothing mechanism based on path structure awareness is used to smooth the path and obtain the smoothed predicted delay value of the path. S2. Based on the smoothed predicted delay value of the path, a composite cost function is used to obtain the comprehensive path cost; based on the comprehensive path cost, the optimal transmission path is determined; based on the optimal transmission path, a comprehensive feedback video quality score calculation formula is introduced to calculate the video quality score; based on the video quality score, the video parameters are adjusted.

[0005] Preferably, S1 specifically includes: In the path structure-aware dynamic smoothing mechanism, the predicted delay value of the path at the current time is used to obtain the predicted delay value of the path at the historical time; based on the predicted delay value of the path at the historical time, variance is calculated to obtain the local fluctuation intensity of the predicted delay.

[0006] Preferably, S1 specifically includes: The predicted delay values ​​of the paths at historical moments are weighted and fused, and then normalized by combining the intensity of local fluctuations in the predicted delay, to obtain the smoothed predicted delay value of the path at the current moment.

[0007] Preferably, S2 specifically includes: Based on the smoothed predicted delay value of the path, calculate the rate of change of the smoothed predicted delay value; based on the sum of the squares of the smoothed predicted delay value and the rate of change of the smoothed predicted delay value, calculate the total intensity of the current delay state of the path.

[0008] Preferably, S2 specifically includes: Based on the total intensity of the current delay state of the path, the path's jump frequency is introduced to construct a composite cost function and calculate the overall cost of the path.

[0009] Preferably, S2 specifically includes: Based on the overall path cost, the path with the lowest overall path cost is selected as the optimal transmission path.

[0010] Preferably, S2 specifically includes: Based on the optimal transmission path, video playback quality feedback parameters are collected; based on the collected video playback quality feedback parameters, a comprehensive feedback video quality score calculation formula is constructed to obtain the video quality score.

[0011] Preferably, S2 specifically includes: Based on the video quality score, it is compared with a set threshold; when the video quality score is greater than or equal to the threshold, the existing video parameters and transmission path remain unchanged; when the video quality score is less than the threshold, the video parameters are adjusted.

[0012] The beneficial effects of the technical solution of the present invention are: 1. By continuously collecting the raw delay values ​​of multiple paths during the data acquisition phase and introducing a composite weighted prediction model, potential risky paths can be identified in advance before the network condition deteriorates significantly, achieving a scheduling effect of "prediction before congestion". Through a dynamic smoothing mechanism based on path structure awareness, the historical delay prediction results and the intensity of local fluctuations in predicted delay are jointly normalized, effectively suppressing non-steady-state interference caused by instantaneous jitter or abnormal sampling, significantly improving the stability and consistency of path evaluation results, and avoiding video stuttering and reconnection problems caused by frequent path switching.

[0013] 2. By constructing a composite cost function that simultaneously includes the smoothed predicted latency value, the rate of change of the smoothed predicted latency value, the frequency of jumps, and the change in path selection probability, the path selection process can achieve a balance between "low latency, low fluctuation, and low switching risk," reducing the probability of image jitter, audio-visual asynchrony, and session interruption caused by frequent path adjustments during video conferencing. Simultaneously, a comprehensive video quality scoring formula based on the client's actual playback status is introduced, allowing adjustment decisions to be directly optimized based on the user's subjective experience, thereby improving the targeting and effectiveness of the adjustment strategy. Attached Figure Description

[0014] Figure 1 This is a flowchart of a low-latency streaming media transmission optimization method for video conferencing based on edge computing, as described in this invention. Detailed Implementation

[0015] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0017] The following description, in conjunction with the accompanying drawings, details a specific scheme for optimizing low-latency streaming media transmission in video conferencing based on edge computing, provided by this invention.

[0018] See attached document Figure 1 The diagram illustrates a flowchart of a low-latency streaming media transmission optimization method for video conferencing based on edge computing, according to an embodiment of the present invention. The method includes the following steps: S1. Based on the collected raw delay value, the predicted delay value of the path is obtained; based on the predicted delay value of the path, a dynamic smoothing mechanism based on path structure awareness is used to smooth the path and obtain the smoothed predicted delay value of the path.

[0019] Network status is collected by edge computing entities with computing, storage, and network scheduling capabilities, such as smart edge routers, 5GMEC terminals, and SD-WAN devices, deployed by technical personnel at the network access edge, close to user terminals or access points. The core metric collected is the performance of each path in continuous operation. The original delay values ​​at each time step, where any one of the original delay values ​​is defined as... ,in, At the current time point, To extend the steps of historical retrospection, Indicates from edge node To edge nodes The path, in the edge computing system, starts from the edge node. To edge nodes Between these, there is an independent data forwarding channel used for streaming media transmission, such as between a video conferencing client device and the target service node. The collected raw latency values ​​are stored in a historical latency database for subsequent calculations.

[0020] To predict video transmission delay trends within a short future time window, a composite weighted prediction model is introduced. The technical objective of this composite weighted prediction model is to predict video transmission delay trends within a specified window length. The path transmission delay level within seconds is quantitatively estimated to provide prior information for subsequent path selection and scheduling algorithms. Unlike traditional simple models based on moving averages or time series extrapolation, the composite weighted prediction model integrates multi-dimensional features such as historical path delay levels and first-order derivative trend enhancement factors to construct a composite weighted prediction structure, as shown in the following expression: in, Indicates the current time From edge nodes To edge nodes The predicted delay value of the path, i.e., the current time. path The predicted delay value; It is a historic moment path The original latency value comes from the historical latency database; It is the step size of the time sliding window, and the index for tracing back the historical sequence. ; These are time-weighted coefficients, determined based on the existing transformer attention mechanism, and their values ​​range from [value range missing]. And the cumulative sum is 1; It is a data processing item, namely the path history delay level, used to logarithmically compress delay values ​​and reduce the impact of extreme values; It is a derived variable, it is a path The rate of change of the original delay value over time is approximately determined by the finite difference method; It is a regulation term, namely the first derivative trend enhancement factor, used for trend regulation to improve the sensitivity to abrupt changes; It is a sliding weighted trend enhancement term; This is a long-term forecast correction factor used in the weighted attenuation term. It is estimated using existing statistical models, such as fluctuation amplitude estimation based on path delay. The reference value range is... ; This is a time-predicted decay rate control factor, manually adjusted according to the network service scenario. In low-fluctuation network environments, such as LANs and enterprise SD-WANs, it is set... This allows forecasts to remain valid for extended periods, particularly in highly volatile environments such as public network conferencing and 5G mobile networks. Rapidly diminishes the impact of long-term forecasts; It is the length of the future time window for predicting the target, which is determined according to specific needs. The reference range is as follows: ; It is an exponentially decaying correction term, used to represent the mild penalty and correction that future uncertainty imposes on the current forecast delay value; This represents the weighted attenuation term.

[0021] Furthermore, to avoid instantaneous non-steady-state jumps in prediction results due to high-frequency fluctuations or sudden values ​​in historical delays, which could severely affect the stability of downstream path cost assessment, a path structure-aware dynamic smoothing mechanism is used to independently smooth the predicted delay values ​​for each path. The core technology of this path structure-aware dynamic smoothing mechanism lies in independently processing the predicted delay value sequence formed on each candidate path. Through weighted fusion of historical delay prediction results, and normalization adjustment based on the local fluctuation intensity of the predicted delay at the corresponding time, the predicted delay values ​​are smoothed and stabilized. The specific formula for the smoothed prediction delay value is as follows: in, It is the current moment. Below, edge nodes To edge nodes The smoothed prediction delay value of the path, i.e., the current time. Next path The smoothed prediction delay value; It is a path In the past The predicted delay value at each time step, i.e., the historical moment. The delayed prediction results; It is the first The weighting coefficients for each time step are used to adjust the contribution relationship between near and far time points. These coefficients are determined based on the existing transformer attention mechanism and their values ​​range from [value range missing]. ; Representing a path In the past The predicted local fluctuation intensity at each time step is taken from a preset local sliding window. The variance of the predicted delay values ​​within the time frame is used to dynamically assess path stability, where This is the window half-width, determined based on specific needs. For example, use 2 for multi-person real-time video conference rooms and 4 for cloud desktops. A reference range is... ; It is the normalization factor, set to This is used to keep the output dimensions consistent with the original prediction delay; It is the prediction delay value after fluctuation normalization, which is used to normalize and adjust the prediction delay value of each historical time point according to its fluctuation degree, and reduce the contribution of large fluctuation values. It is a weighted normalized fusion term used to represent the path. in the past The predicted delay value is the weighted average of the predicted delay value after weight adjustment and fluctuation normalization within each time step.

[0022] S2. Based on the smoothed predicted delay value of the path, a composite cost function is used to obtain the comprehensive path cost; based on the comprehensive path cost, the optimal transmission path is determined; based on the optimal transmission path, a comprehensive feedback video quality score calculation formula is introduced to calculate the video quality score; based on the video quality score, the video parameters are adjusted.

[0023] After smoothing, in order to select the path with the lowest overall cost from all paths as the main transmission path for the video stream, a composite cost function is used to evaluate the performance of each path. The composite cost function is as follows: in, Indicates from edge node To edge nodes The total cost of the path, i.e., the path The overall cost is used to rank the paths based on their merits. Representing a path The rate of change of the smoothed prediction delay value over time; the larger the value, the higher the volatility, which is calculated by the first-order difference method. For path The number of times a path is switched (enabled or disabled) within the last 5 scheduling cycles, i.e., the hop frequency, is an indicator of path stability. The scheduling cycle is the time required for path reassessment. This is the path switching penalty adjustment coefficient, used to adjust the intensity of the penalty for decision instability. It is determined using cross-validation, and the reference value range is [value missing]. ; and For the current moment Compared to the previous moment path The estimated probability of being selected is obtained through continuous... The path selection frequency within a sliding window at each time step is determined statistically, with a reference range of values. ; Used to reflect whether the route selection changes frequently; The time length between adjacent path scheduling cycles is determined based on scheduling parameters preset by technical professionals, such as path scheduling frequency, scheduling triggering conditions, and minimum / maximum interval constraints. It is a path The total intensity of the current delay state; It is a penalty adjustment based on path state energy and historical switching frequency; It represents the strength of the path switching trend, used to suppress frequent path switching behavior, thereby reducing stuttering and reconnection in video conferencing.

[0024] Furthermore, based on the composite cost function, the path with the minimum overall cost is selected. From edge nodes To edge nodes The optimal transmission path is selected to enable streaming media transmission.

[0025] Once a path is selected and a communication connection is established, the video stream is transmitted to the client along that path. The client continuously collects and reports video playback quality feedback parameters, including packet loss rate, at the receiving end. Current remaining capacity of the decoding buffer Total buffer capacity Frame rate change rate After receiving video playback quality feedback parameters, the edge node uses a comprehensive feedback video quality score calculation formula to calculate the video quality score. As a quantitative evaluation indicator of the current video playback status, it takes the following form: in, It is the video quality score reported by the client at the current time, which is used to drive parameter adjustments on the sending end; The average packet loss rate is reported by the client. It is the arctangent function, used to construct a nonlinear response mechanism to the critical state of a buffer. Indicates frame rate; The frame rate change rate is measured by the client player. To enhance the penalty for drastic jumps; It is the remaining capacity of the buffer, that is, the remaining space in the current client player's buffer, which is uploaded and fed back by the client; This is the total buffer capacity, used to represent the player's maximum buffer capacity; It is a constant for preventing extremely small positive numbers from having a denominator of 0, such as ; This is a packet loss weighting factor used to control the impact of packet loss on video quality scoring and adjust sensitivity to instability. The reference value range is... ; This is a buffer weighting factor used to control the impact of remaining buffer capacity on video quality scoring, reflecting sensitivity to buffer changes. The reference value range is... ; This is the frame rate fluctuation penalty coefficient, used to control the severity of the penalty for frame rate fluctuations in the video quality score. A larger value indicates less tolerance for jitter. The reference value range is... ; , , All were determined using the grid search method; It represents the positive contribution of network transmission reliability to the subjective experience of video conferencing, indicating how much effective carrying capacity the network can still provide for video conferencing under the current packet loss level; It is the playback-side time redundancy capability, also known as buffer safety margin, which is a positive contribution to the continuity of video. It is used to indicate how long the player can continue playing with the existing buffer under the current network conditions. It represents the severity of the penalty to the user experience caused by the dynamic instability of video playback, and is used to measure whether the video output state is in a state of drastic adjustment or oscillation.

[0026] Ultimately, The value and the threshold preset by professional technicians When comparing, When this occurs, it indicates that there are no obvious frame drops, stuttering, or insufficient buffering during the current video playback process, and the transmission status is good. Maintaining existing video parameters such as video encoding bitrate, frame rate, and keyframe spacing, as well as the transmission path, the system enters the monitoring and maintenance state. When a quality degradation state is detected, a preset adaptive adjustment mechanism is immediately triggered. Based on the current network carrying capacity, one or more adjustment strategies are executed, including: reducing the video encoding bitrate to alleviate bandwidth pressure, dynamically compressing the frame rate to reduce data load, shortening the key frame interval to improve recovery capability, or enabling a local frame dropping strategy to ensure smooth playback. The adjusted video parameters will take effect in the next transmission cycle, thereby achieving rapid self-repair of video quality in weak network or abnormal environments.

[0027] In summary, a method for optimizing low-latency streaming media transmission in video conferencing based on edge computing has been developed.

[0028] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0029] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0030] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for optimizing low-latency streaming media transmission in video conferencing based on edge computing, characterized in that, Includes the following steps: S1. Based on the collected raw delay value, the predicted delay value of the path is obtained; based on the predicted delay value of the path, a dynamic smoothing mechanism based on path structure awareness is used to smooth the path and obtain the smoothed predicted delay value of the path. S2. Based on the smoothed predicted delay value of the path, a composite cost function is used to obtain the comprehensive path cost; based on the comprehensive path cost, the optimal transmission path is determined; based on the optimal transmission path, a comprehensive feedback video quality score calculation formula is introduced to calculate the video quality score; Adjust video parameters based on video quality score.

2. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 1, characterized in that, S1 specifically includes: In the path structure-aware dynamic smoothing mechanism, the predicted delay value of the path at the current time is used to obtain the predicted delay value of the path at the historical time; based on the predicted delay value of the path at the historical time, variance is calculated to obtain the local fluctuation intensity of the predicted delay.

3. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 2, characterized in that, S1 specifically includes: The predicted delay values ​​of the paths at historical moments are weighted and fused, and then normalized by combining the intensity of local fluctuations in the predicted delay, to obtain the smoothed predicted delay value of the path at the current moment.

4. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 1, characterized in that, S2 specifically includes: Based on the smoothed predicted delay value of the path, calculate the rate of change of the smoothed predicted delay value; based on the sum of the squares of the smoothed predicted delay value and the rate of change of the smoothed predicted delay value, calculate the total intensity of the current delay state of the path.

5. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 4, characterized in that, S2 specifically includes: Based on the total intensity of the current delay state of the path, the path's jump frequency is introduced to construct a composite cost function and calculate the overall cost of the path.

6. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 5, characterized in that, S2 specifically includes: Based on the overall path cost, the path with the lowest overall path cost is selected as the optimal transmission path.

7. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 6, characterized in that, S2 specifically includes: Based on the optimal transmission path, video playback quality feedback parameters are collected; based on the collected video playback quality feedback parameters, a comprehensive feedback video quality score calculation formula is constructed to obtain the video quality score.

8. The method for optimizing low-latency streaming media transmission in video conferencing based on edge computing according to claim 7, characterized in that, S2 specifically includes: Based on the video quality score, it is compared with a set threshold; when the video quality score is greater than or equal to the threshold, the existing video parameters and transmission path remain unchanged; when the video quality score is less than the threshold, the video parameters are adjusted.

Images