Bandwidth prediction method based on frame-level burst and real-time communication congestion control method and system

Abstract

The invention provides a bandwidth prediction method based on frame level burst and a real-time communication congestion control method and system. Through the bandwidth prediction method based on frame-level burst, effective prediction of the network bandwidth under the condition of insufficient data is realized. A congestion control system for real-time audio and video transmission is composed of a network bandwidth sampling module, a general video frame size modeling module, a real-time video frame delay estimation module and a fast convergence target code rate decision module, so that the influence of code rate fluctuation and network congestion on the audio and video transmission quality is remarkably reduced; therefore, the playing experience of the user is greatly improved. Therefore, the bandwidth prediction method based on frame level burst and the real-time communication congestion control system provided by the invention can effectively optimize the congestion control of audio and video transmission application, improve the video code rate, avoid network congestion and remarkably improve the playing experience of a user while ensuring the transmission quality.

Description

Technical Field

[0001] The present invention belongs to the field of network real-time communication, and in particular to the scope of audio and video transmission technology. Specifically, the present invention proposes an innovative bandwidth prediction method based on frame-level burst characteristics, and a set of real-time communication congestion control methods and systems matched therewith. The method aims to improve the transmission efficiency and stability of audio and video data in real-time communication. Background Art

[0002] As the core of network communication, real-time audio and video transmission technology is becoming increasingly important in today's society. With the deepening of digital transformation, the demand for application scenarios such as distance education, online medical care, and interactive live broadcast has surged. Real-time audio and video transmission technology has become a key link to connect users and transmit information. Given the strict real-time requirements of real-time audio and video transmission, the transmission process relies on efficient and low-latency congestion control algorithms. Such algorithms can dynamically adjust the sending rate or congestion window size of the sender based on the real-time feedback information of the receiver, and synchronously adjust the encoding bit rate of the encoder. Its purpose is to maximize the bit rate of video transmission while avoiding network congestion, thereby ensuring the optimal user playback experience of upper-layer applications.

[0003] In traditional real-time audio and video transmission protocols, congestion control algorithms are mostly designed for the network. They inherit the design concept of traditional TCP congestion control algorithms and assume that upper-layer applications can continuously provide sufficient data, thereby allowing data in the send buffer to continue to accumulate. These algorithms usually rely on congestion indicators such as network delay, network queue length, or packet loss to adjust the transmission rate. When a congestion signal is detected, the algorithm will consider the network to be congested and reduce the transmission rate accordingly; conversely, it will increase the rate when there is no congestion signal. This design mode based on congestion signal response can usually effectively detect network bandwidth and prevent network congestion in non-real-time application scenarios such as file downloads and video on demand.

[0004] However, in the field of real-time audio and video transmission, the assumption that data is always sufficient is no longer applicable. This is because video frames are encoded in real time, and their frame delay is calculated from the moment the encoding is completed. In the application scenario of real-time audio and video transmission, the server encodes the original video in real time and sends the encoded video data packets to the sending buffer. Due to the spatial and temporal complexity of the video content itself, the actual bit rate output by the encoder tends to fluctuate around the target bit rate. When the actual bit rate generated by the encoder exceeds the sending rate set by congestion control, the newly encoded video frames will queue in the sending buffer, resulting in an increase in frame delay. On the other hand, when the actual bit rate is lower than the sending rate of congestion control and the sending buffer is empty, the congestion control algorithm lacks sufficient data to accurately estimate the available bandwidth of the network. This situation may cause the algorithm to misjudge the bandwidth, affecting the smoothness and stability of real-time audio and video transmission.

[0005] Traditional network-oriented congestion control algorithms usually use heuristic methods to overcome the above problems. Some common supplementary strategies include: using heuristic rules to judge the adequacy of upper-layer data and stopping bandwidth detection when data is insufficient; actively sending filler packets to maintain the sending rate when data is short; when the sending buffer is full, actively lowering the target bit rate of the encoder to reduce the amount of data generated by the upper-layer application. However, these methods still face some limitations, such as high misjudgment rate, unnecessary consumption of traffic, and frequent fluctuations in the target bit rate. These shortcomings may affect the efficiency and quality of real-time audio and video transmission, thereby negatively affecting the user experience. Therefore, improving these algorithms to make them more adaptable to the needs of real-time audio and video transmission is an important research direction in the current field of network communication technology. Summary of the invention

[0006] The present invention aims to provide an innovative bandwidth prediction method based on frame-level burst and a real-time communication congestion control system, aiming to optimize the congestion control strategy of audio and video transmission applications, thereby improving the video transmission bit rate, preventing network congestion, and significantly enhancing the user's playback experience.

[0007] An innovation of the present invention is to propose a universal bandwidth prediction method based on frame-level burst, which is particularly suitable for scenarios with insufficient data to accurately estimate network bandwidth. The method comprises the following steps:

[0008] Based on the principle of allocating network bandwidth resources to different flows according to the rate, a general network competition model is constructed;

[0009] According to the network competition model, in view of the fact that data is generated by frames in audio and video transmission, the frame-level burst technology is used to determine the upper limit of network bandwidth by analyzing the receiving rate of video frames.

[0010] The frame sending rate is adjusted in time according to the network competition model and the network bandwidth upper limit, and then the average rate of the network background flow is derived.

[0011] Another innovation of the present invention is to propose a congestion control method suitable for real-time audio and video transmission scenarios, comprising the following steps:

[0012] Using the above-mentioned bandwidth prediction method based on frame-level burst, the network bandwidth upper limit and the average sending rate of background traffic are obtained;

[0013] Using the frame size data of recent video frame samples, a mapping relationship from target bit rate and picture complexity to video frame size distribution is established;

[0014] Calculate the average frame delay after any time period by combining the video frame size distribution, network bandwidth upper limit and average background traffic sending rate;

[0015] According to the target utility function related to the target bit rate and the average frame delay, the optimal bit rate to achieve the maximum utility is calculated, and a target bit rate decision that can converge quickly is achieved.

[0016] Another innovation of the present invention is to provide a congestion control system suitable for real-time audio and video transmission scenarios, which consists of the following modules:

[0017] Network bandwidth sampling module: uses the bandwidth prediction method based on frame-level burst to obtain the network bandwidth upper limit and the average rate of background flow;

[0018] General video frame size modeling module: By analyzing the frame size data of a small number of recent video frame samples, a mapping relationship between target bit rate and picture complexity to video frame distribution is established;

[0019] Real-time video frame delay estimation module: combines the video frame size distribution and the parameters provided by the network bandwidth sampling module to calculate the average frame delay after any time period;

[0020] Target bitrate decision module: Based on a custom target utility function, which is related to the target bitrate and network delay, the optimal bitrate for achieving maximum utility is directly derived, thereby achieving fast convergence of target bitrate decisions.

[0021] The present invention provides a bandwidth prediction method and congestion control method and system based on frame-level burst for real-time audio and video transmission. The bandwidth prediction method based on frame-level burst is used to achieve effective prediction of network bandwidth under insufficient data conditions. The congestion control system for real-time audio and video transmission composed of a network bandwidth sampling module, a general video frame size modeling module, a real-time video frame delay estimation module and a fast-converging target bit rate decision module significantly reduces the impact of bit rate fluctuations and network congestion on the quality of audio and video transmission, thereby greatly improving the user's playback experience. Therefore, the bandwidth prediction method based on frame-level burst and the real-time communication congestion control method and system provided by the present invention can effectively optimize the congestion control of audio and video transmission applications while ensuring transmission quality, improve video bit rate, avoid network congestion, and significantly improve the user's playback experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart of a real-time audio and video bandwidth prediction method provided by an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of the structure of a congestion control system for a real-time audio and video transmission scenario provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0025] The embodiment of the present invention aims at the problem that the bandwidth estimation method of congestion control in the existing real-time audio and video transmission protocol is not suitable for the bit rate fluctuation of real-time audio and video applications. Firstly, a bandwidth prediction method based on frame-level burst is proposed. Figure 1 The flowchart of the bandwidth prediction method based on frame-level burst provided in this embodiment, the specific steps of the method are as follows:

[0026] Step 101: Establish a general network competition model that allocates network bandwidth resources to different flows according to rates.

[0027] For a connection of an audio or video transmission application, the amount of available bandwidth is affected by the network bottleneck bandwidth capacity and background traffic.

[0028] During the operation of audio and video transmission applications, the transmitted data will pass through multiple network nodes on the Internet and be forwarded across multiple links. In this transmission path, a specific link or node with the lowest bandwidth capacity will constitute the limiting point of the entire data transmission link, and the bandwidth of this link or node is defined as the "bottleneck bandwidth". The capacity of this "bottleneck bandwidth" determines the bandwidth upper limit of the entire transmission link.

[0029] In this embodiment, it is assumed that the bottleneck device meets the first-come-first-served service condition, and the bottleneck bandwidth resources are allocated roughly according to the rate at which each connection reaches the bottleneck. In this embodiment, a connection path for an audio and video transmission application is abstracted into a network model with a bottleneck bandwidth capacity of C and an average background traffic sending rate of B. The connection sends data at a rate S at the sending end, and a receiving rate R can be measured at the receiving end. According to the above-mentioned network resource allocation principle, the formula (1) expressing the relationship between the sending rate and the receiving rate of the connection can be written, that is, the network competition model:

[0030]

[0031] Step 102: Adapting to the characteristic that data of audio and video transmission is generated at frame granularity, a frame-level burst method is used to obtain the bandwidth upper limit of the network according to the receiving rate of the video frame.

[0032] According to the network competition model of step 101, the greater the sending rate, the closer the receiving rate is to the upper limit of the network bandwidth. When the sending rate is infinite, the receiving rate is approximately equal to the network bandwidth, so the receiving rate of the burst frame can be used to estimate the upper limit of the network bandwidth.

[0033] In this embodiment, the receiving rate of the burst frame is calculated using the receiving time difference between the first and last packets of the frame and the frame size. For a burst frame with n packets, the receiving timestamp of the first packet is rt1, and the receiving timestamp of the last packet is rt n , the packet size is P. Then the calculation formula can be written as:

[0034]

[0035] Step 103: According to the network competition model, the frame sending rate is appropriately adjusted to derive the average rate of the network background flow.

[0036] For an audio or video connection, measuring the network bandwidth limit alone is not sufficient to determine the target bitrate, because background traffic in the network will affect the actual available bandwidth of the application.

[0037] According to the network competition model in step 101, when the sending rate S, receiving rate R, and bandwidth upper limit C are known, the background traffic can be derived by formula (3):

[0038]

[0039] In this embodiment, in order to derive the average sending rate B of the background traffic, the frame sending rate is appropriately adjusted to C to ensure competition with the background flow, so that the sending rate, receiving rate and bandwidth upper limit all become measurable or known quantities, and the average sending rate B of the background traffic is derived.

[0040] All calculations in this embodiment are obtained from the information of a single frame, and the information of the overall transmission rate of the audio and video connection is not used, so it can work under any data conditions, thereby achieving effective detection of network bandwidth under insufficient data conditions.

[0041] Figure 2 A schematic diagram of the structure of a congestion control system adapted to real-time audio and video transmission scenarios provided by an embodiment of the present invention, the system comprising:

[0042] The network bandwidth sampling module 11 is used to obtain the network bandwidth upper limit and the average sending rate of the background traffic by using a bandwidth prediction method based on frame-level bursts;

[0043] The general video frame size modeling module 12 is used to establish a mapping from target bit rate and picture complexity to video frame size distribution using frame size data of a small amount of recent video frame samples.

[0044] The real-time video frame delay estimation module 13 is used to derive the average frame delay after any time period based on the given video frame size distribution, the network bandwidth upper limit provided by the network bandwidth sampling module, and the average background traffic sending rate.

[0045] The target bit rate decision module 14 is used to directly derive the optimal bit rate for achieving maximum utility based on a user-defined target utility function related to the target bit rate and the average frame delay, so as to achieve a target bit rate decision that can converge quickly.

[0046] The technical principle of the network bandwidth sampling module 11 is Figure 1 In addition to the above technical principles, the network bandwidth sampling module 11 also needs to decide which frames are directly burst to the network, which frames adjust the sending rate to measure the background traffic, and how many frames are burst at one time.

[0047] In this embodiment, a frame is preferably divided into two equal lengths, the sending rate of the first half of the frame is adjusted to the estimated bandwidth size, and the second half is directly burst to the network, and the background traffic and bandwidth limit are calculated using the above technical principles. In addition, in order to prevent excessive data from bursting at one time when sending key frames, the length of the second half is limited, and in low bit rate scenarios, the method of using burst odd-numbered frames to adjust even-numbered frames is changed.

[0048] The general video frame size modeling module 12 is used to establish a mapping from target bit rate and picture complexity to video frame size distribution using frame size data of a small amount of recent video frame samples. The technical principle is as follows:

[0049] In this embodiment, the modeled video frame size distribution is a normal distribution, and the mean and variance of the frame size distribution are maintained respectively.

[0050] Specifically, each time the transmitter completes the encoding of a frame of video content, the modeling module 12 collects four pieces of information related to the video frame: frame size F, encoder target bit rate T, frame time complexity TI, and frame space complexity SI. For the mean μ of the normal distribution model, this embodiment uses the parameter W∈R 1×3 The first-order three-variable linear function fitting is written as formula (4):

[0051] μ=W×[T,TI,SI] (4)

[0052] The samples used for the parameters in the fitting formula (4) are the most recently collected 30 frames of samples, and the observed value of μ is represented by the frame size F.

[0053] For the variance σ of the normal distribution model, this embodiment uses an exponential moving average algorithm with a parameter a to maintain it, and writes formula (5):

[0054]

[0055] Among them, μ, σ i They represent the mean and variance of the normal distribution of the frame size, Fi represents the frame size of the i-th frame, and α is the hyperparameter used for the exponential moving average.

[0056] The real-time video frame delay estimation module 13 is used to derive the average frame delay after any time period based on the given video frame size distribution and the network bandwidth upper limit and background flow average rate provided by the network bandwidth sampling module. The specific technical principle is as follows:

[0057] In this embodiment, the bandwidth and background traffic information measured by the network bandwidth sampling module 11 and the video distribution information modeled by the general video frame size modeling module 12 are used to model the relationship between the first packet queuing delay of two adjacent frames, and derive the queuing delay after any time, thereby calculating the frame delay. When a frame is sent to the network, before the next frame arrives, the amount of data forwarded by the bottleneck network node affects the queuing delay of the first packet of the next frame. The frame F sent at time t-Δt is recorded as t-Δt , the queuing delay of the bottleneck network node when it reaches the bottleneck network node is recorded as q t-Δt After that, all the data packets of the frame arrive at the bottleneck network node in a very short time. The frame F sent at time tt The queue delay at its arrival is denoted as q t In the Δt time, except for the last frame arriving at the bottleneck network node, the background traffic continues to send data to the bottleneck network node at a sending rate of B. The total data arriving at the bottleneck network node is F t-Δt +B×Δt. If the queue is not emptied within Δt, the bottleneck network node will forward C×Δt of data, otherwise the queue delay will be 0. Therefore, the change relationship of the queue delay within Δt can be written as formula (6):

[0058]

[0059] In specific implementation, since the frame does not wait additionally in the sending buffer, the queuing delay of the frame head packet directly reflects the frame delay. Therefore, this embodiment uses the queuing delay of the frame head packet to represent the estimated value of the frame delay of the frame. When the sending end obtains the frame delay size of a frame through network feedback, the frame delay size of any subsequent frame can be iteratively derived using formula (6).

[0060] The fast-converging target bit rate decision module 14 is used to directly derive the optimal bit rate for achieving maximum utility based on a user-defined target utility function related to the target bit rate and average frame delay, and to achieve a target bit rate decision that can converge quickly. The technical principles are as follows:

[0061] In this embodiment, the target utility function is defined as formula (7) related to the target bit rate T and the average frame delay D:

[0062] U(T)=log(αT+β)-wD (7)

[0063] In this embodiment, the frame delay D can be expressed by the average of the queuing delay q over a period of time. After calculation by the real-time video frame delay estimation module 13, the above formula has been completely determined numerically. Since the utility function tends to increase first and then decrease with the increase of the target bit rate, the maximum value of the target bit rate is formula (8):

[0064] T = argmaxU(T)

[0065] In specific implementation, since the derivation of the frame delay D is relatively complex, the direct derivation method is not applicable. This embodiment uses the traversal solution to obtain the optimal target bit rate. The current target bit rate is T i-1 , then calculate the target bit rate set {T i-1 ×0.5,T i-1 ×0.5+50kbps,…,T i-1×1.5}, that is, from half of the current target bit rate to 1.5 times of the current target bit rate, the utility function values ​​corresponding to all target bit rates in the target bit rate set covered by the step size of 50 kbps are taken, and the bit rate with the largest utility value is selected as the target bit rate.

[0066] The congestion control system adapted to the real-time audio and video transmission scenario of the embodiment of the present invention is used to maximize the bandwidth information measured by the bandwidth prediction method based on frame-level burst, while considering the impact of bit rate fluctuations on audio and video transmission, and decide the optimal target bit rate corresponding to the target utility, so as to improve the transmission quality and optimize the user playback experience.

[0067] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, and the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the parameters mentioned in the embodiments can be replaced with other parameters within a reasonable range according to the actual application.

[0068] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0069] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0070] The embodiment of the present invention applies a bandwidth prediction method based on frame-level bursts, establishes a network competition model, divides a single frame into two parts or an odd-even frame grouping scheme, measures the network bandwidth capacity upper limit and background traffic respectively, and implements a bandwidth detection method that does not rely on sufficient data; uses normal distribution, linear model, etc. to model the frame size distribution; and uses the utility function shown in formula (7) to achieve a fast convergence target bit rate decision. Among them, using other frame grouping schemes, frame size distribution models and fitting models and utility functions can also achieve the purpose of model construction and decision optimization, but the algorithm used in the embodiment of the present invention has the best experimental effect.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A bandwidth prediction method based on frame-level burst, characterized in that: The following steps are involved: Based on the principle of allocating network bandwidth resources to different flows according to the rate, a network competition model is constructed; According to the network competition model, the frame-level burst technology is used to determine the network bandwidth upper limit by analyzing the reception rate of video frames; The frame sending rate is adjusted according to the network competition model and the network bandwidth upper limit, and then the average sending rate of the background traffic is derived.

2. The method according to claim 1, characterized in that The network competition model is: Among them, C is the upper limit of the network bandwidth for audio and video transmission, B is the average sending rate of background traffic, S is the sending rate of the sender, and R is the receiving rate of the receiver.

3. The method according to claim 2, characterized in that The frame-level burst technology uses the receiving rate of burst frames to estimate the upper limit of network bandwidth, where the receiving rate of burst frames is calculated using the receiving time difference between the first packet and the last packet of the burst frame and the frame size, and the calculation formula is: Where n is the number of packets in a burst frame, rt1 is the receiving timestamp of the first packet of the burst frame, and rt n It is the receiving timestamp of the tail packet of the burst frame, and the packet size is P.

4. The method according to claim 3, characterized in that The adjustment of the frame sending rate is to adjust the frame sending rate to C, to ensure competition with the background flow, so that the sending rate S, the receiving rate R and the network bandwidth upper limit C are all measurable or known quantities, and then the background flow average sending rate B is derived:

5. A congestion control method suitable for real-time audio and video transmission scenarios, characterized in that: The following steps are involved: Using the bandwidth prediction method based on frame-level bursts as described in any one of claims 1 to 4, the network bandwidth upper limit and the average background traffic sending rate are obtained; Using the frame size data of recent video frame samples, a mapping relationship from target bit rate and picture complexity to video frame size distribution is established; Calculate the average frame delay after any time period by combining the video frame size distribution, network bandwidth upper limit and average background traffic sending rate; According to the target utility function related to the target bit rate and the average frame delay, the optimal bit rate to achieve the maximum utility is calculated, and a target bit rate decision that can converge quickly is achieved.

6. The method according to claim 5, characterized in that The obtaining of the network bandwidth upper limit and the background traffic average sending rate includes: A frame is divided into two equal lengths, the sending rate of the first half of the frame is adjusted to the estimated bandwidth size, and the second half is directly burst to the network, and then the network bandwidth limit and the average sending rate of background traffic are calculated; The upper limit of the length of the second half is limited, and in low bit rate scenarios, the frames with burst odd frame numbers are changed to adjust the frames with even frame numbers.

7. The method according to claim 5, characterized in that The use of frame size data of recent video frame samples to establish a mapping relationship from target bit rate and picture complexity to video frame size distribution includes: The modeled video frame size distribution is a normal distribution, and the mean and variance of the frame size distribution are maintained respectively; Each time the sender completes the encoding of a video frame, it collects four pieces of information related to the video frame: frame size F, encoder target bit rate T, frame temporal complexity TI, and frame spatial complexity SI; For the mean μ of the normal distribution model, use the parameter W∈R 1×3 The first-order three-variable linear function is fitted, and the fitting formula is μ = W × [T, TI, SI]. The samples used by the parameters in the fitting formula are the multi-frame samples collected recently, and the observed value of μ is represented by the frame size F; For the variance σ of the normal distribution model, an exponential moving average algorithm with parameter a is used to maintain it.

8. The method according to claim 5, characterized in that The calculation of the average frame delay after any time period by combining the video frame size distribution, the network bandwidth upper limit and the average background traffic sending rate includes: The frame F sent at time t-Δt is t-Δt , the queuing delay of the bottleneck network node when it reaches the bottleneck network node is recorded as q t-Δt ,After that, all the data packets of the frame arrive at the bottleneck network node in a very short time; The frame F sent at time t t The queue delay at its arrival is denoted as q t In the Δt time, except for the last frame arriving at the bottleneck network node, the background traffic continues to send data to the bottleneck network node at a sending rate of B. The total data arriving at the bottleneck network node is F t-Δt +B×Δt; If the queue has not been emptied within Δt, the bottleneck network node forwards c×Δt of data, otherwise the queue delay will be 0. Therefore, the change relationship of the queue delay within Δt is:

9. The method according to claim 5, characterized in that The target utility function is defined as a formula related to the target bit rate T and the average frame delay D: U(T)=log(αT+β)-wD.

10. A congestion control system suitable for real-time audio and video transmission scenarios, characterized in that: include: A network bandwidth sampling module, used to obtain the network bandwidth upper limit and the average sending rate of the background traffic by using the bandwidth prediction method based on the frame-level burst according to any one of claims 1 to 4; A general video frame size modeling module is used to establish a mapping relationship from target bit rate and picture complexity to video frame size distribution using frame size data of a small number of recent video frame samples; The real-time video frame delay estimation module is used to calculate the average frame delay after any time period by combining the video frame size distribution, network bandwidth upper limit and background traffic average sending rate. The target bit rate decision module is used to calculate the optimal bit rate for achieving maximum utility based on the target utility function related to the target bit rate and the average frame delay, so as to achieve a target bit rate decision that can converge quickly.

Images