CN101938341B - Cross-node controlled online video stream selective retransmission method - Google Patents

Abstract

The invention relates to a cross-node controlled online video stream selective retransmission method, which includes the following steps: a single video stream encoded online enters the sending end, and is divided into DFSs one by one according to the start-up and transmission delay restrictions, as The basic object of selective retransmission; the sender, as the first communication node, determines the importance level of IP packets in DFS according to the importance measure of fault tolerance, and provides selective retransmission of IP packets through packet scheduling based on the importance of fault tolerance ; Based on the importance level of the IP packet, the base station, as the second communication node, provides selective retransmission of the wireless link unit through an ARQ method with decreasing priority. Under the limitation of delay and bandwidth conditions, the present invention organically combines the priority transmission mechanism between the main communication nodes, realizes the importance division and selective retransmission of the wireless link unit level, and can enhance the error of online video stream Control performance to effectively improve the reconstruction quality of the received video.

Description

A method for selective retransmission of online video stream with cross-node control

technical field

The invention relates to the field of image communication, in particular to a cross-node controlled online video stream selective retransmission method.

Background technique

With the development of image communication and broadband cellular communication technology (such as the evolving B3G/4G cellular communication), cellular video streaming applications based on the Internet Protocol (Internet Protocol, referred to as "IP") are gradually emerging. Different from the on-demand streaming transmission based on offline video encoding, the online application mode of cellular video streaming (referred to as "online video streaming") is under the limit of start-up delay acceptable to a single user, the sending end encodes a single video online while the receiving end Replay for a limited time. The development of advanced video coding and a new generation of broadband wireless mobile communication has provided the possibility for such new applications, but its error control faces greater technical challenges. For a two-way cellular communication network that allows retransmission, the entire system can be divided into two parts: the wireless access network and the wired core network. The wireless link layer needs to further divide IP packets into smaller wireless link units. For online video streaming, the stream primitive is the smallest transmission unit that can be independently retransmitted between communication nodes. Specifically, the IP packet is the stream primitive of the wired core network, and the wireless link unit is the stream primitive of the wireless access network. Yuan. With the increasingly complex space-time prediction technology in video coding, the sensitivity of video streams to the loss of different stream primitives is quite different, resulting in different degrees of video degradation. In a cellular communication system, the radio access network part is the bottleneck for the reliable transmission of video streams because it has lower bandwidth and higher bit error rate than the wired core network. After the video stream passes through the wireless access network, there is often a sudden loss of wireless link units, which often causes a serious decline in the quality of the reconstructed video. Therefore, the error control of online video stream is a key technical problem that must be solved. As we all know, Automatic Repeat Request (Automatic Repeat Request, referred to as "ARQ") and forward error correction (Forward Error Correction, referred to as "FEC") are two basic error control methods. In the cellular communication system, FEC is suitable for the error control of the physical layer. Currently, RS codes, LDPC codes, and Turbo codes are commonly used. After the physical layer FEC provides basic error control capabilities, the ARQ method can flexibly enhance the error control performance of online video streams at the data link layer or network layer.

For online video streams that allow a certain delay, selective retransmission technology is an effective measure to improve its error control performance. Efficient distribution for overall optimization of reconstruction quality. Video streams usually need to introduce a Group Of Pictures ("GOP") structure to allow random access, error control, and multimedia synchronization. In GOP-level importance classification, I frames, P frames, and B frames usually have decreasing transmission priorities. Furthermore, a scalable video encoder can generate more levels of transmission priorities according to prediction relationships. However, the existing prioritization method does not consider the transmission distortion, and cannot realize the importance classification of the radio link units. Therefore, according to Shannon's related theory of source-channel separation, the key to selective retransmission is to efficiently estimate GOP-level transmission distortion to evaluate the fault-tolerant importance of current flow primitives in a basic object. The GOP-level transmission distortion of a stream primitive can be calculated off-line using re-encoded analysis-synthesis technology, but due to computational constraints and real-time constraints, this analysis-synthesis technology is difficult to apply to online video streams. Existing selective retransmissions can only provide very limited types of importance for fault tolerance. How to estimate the GOP-level transmission distortion of each flow primitive online to realize the importance division and selective retransmission of radio link unit accuracy, Have not seen relevant technical literature at present. As a typical ARQ mechanism, hybrid automatic retransmission request (Hybrid-ARQ, referred to as "HARQ") technology has been used in 3G high-speed packet access (High-Speed Packet Access, referred to as "HSPA") network to wireless link unit error Control, but the data link layer mechanism of HARQ does not consider the difference of video content, and cannot distinguish the fault tolerance importance of different wireless link units.

After searching the prior art of the same kind, it was found that the patent "Wireless Video Stream Using Single-Layer Coding and Prioritization Streaming" (publication number: CN101073268; classification number: H04N7) proposes to assign a priority level to each coded video frame , and proposes to selectively send some or all video frames according to their levels based on bandwidth constraints. This patent only provides a prioritization and transmission method of the application layer, which cannot be applied to the selective retransmission of the network layer or the data link layer. The priority level only depends on the number of frames in a GOP, and cannot Applied to small stream primitives such as wireless link units; this patent only considers the decoding dependency between the front and rear frames in the GOP, and does not consider the transmission distortion of the reconstructed video. The non-stationary nature of the video content makes the above method inaccurate Reflects the actual priority of encoded frames.

Research in recent years has shown that the cross-layer (Cross-layer) design for wireless video breaks the isolation between the layers of the communication protocol stack to a certain extent, and extracts, analyzes and shares information on the parameters and status information of each layer. The video content is optimized for judgment and policy adjustment to realize the optimal allocation of resources and self-adaptation to channel fluctuations, so as to achieve the best overall performance. The existing cross-layer retransmission technology is still limited to two nodes in the communication network, and does not consider the joint optimization mechanism between more communication nodes, it is difficult to adapt to non-stationary video content and burst errors in a single video stream channel status. In a cellular network shared by multiple users, the base station often needs to process multiple video streams at the same time. Therefore, for a single video stream, it only supports low-complexity importance division and selective retransmission. The base station hopes that the sender can provide some kind of cooperation Processing mechanism to help it reduce the amount of calculation. Therefore, in a cellular communication system, it is necessary to provide a priority retransmission method that combines source and channel characteristics for online video streams, so as to improve the reconstruction quality of received videos.

Contents of the invention

The technical problem to be solved by the present invention is to provide a cross-node controlled online video stream selective retransmission method, which can enhance the error control performance of the live video stream, thereby effectively improving the reconstruction quality of the received video.

The technical solution adopted by the present invention to solve the technical problem is to provide a cross-node controlled online video stream selective retransmission method, comprising the following steps:

(1) According to the start-up and transmission delay constraints, the sender divides the online encoded single video stream into one after another delay-sensitive frame set (Delay-sensitive Frame Set, referred to as "DFS") as the basis for selective retransmission object;

(2) The sender, as the first communication node, first performs packet scheduling based on the importance of fault tolerance to provide DFS with selective retransmission of IP packets;

(3) Based on the importance level of the IP packet, the base station, as the second communication node, implements an automatic repeat request mode with decreasing priority to provide the DFS with selective retransmission of the wireless link unit.

The DFS is a video frame segment divided for a single video stream selection for retransmission, wherein each delay-constrained frame set contains an equal number of video frames; at the sending end, each video frame is encoded as P IP packets, P≥1; at the base station, each IP packet is further encapsulated into K wireless link units with a fixed length, K≥1.

The packet scheduling based on fault-tolerant importance in the described step (2) comprises the following steps:

Obtain and store sideband information; said sideband information includes packaging method, error compensation method, DFS length and real-time encoding result;

Collect IP packets from the same DFS into an encoding queue Q in encoding order;

Determine the relative importance of each IP packet in the coding queue Q to which it belongs according to the fault-tolerant importance measure, and give each IP packet a unique importance level;

Reorder all the IP packets in the encoding queue Q in the order of decreasing importance levels, thus obtaining a new rearrangement queue Q';

Allocate increasing RTP serial numbers to the IP packets in the reordering queue in descending order of importance, and send out the reordering queue Q' in the same scheduling time slot in increasing order of RTP serial numbers.

The base station in the described step (3) comprises a network layer agent and a link layer agent; the described automatic repeat request mode of descending priority comprises the following steps:

Rearrangement queue Q' is received by the network layer agent, and the order of IP packets rearranged is restored according to the real-time transport protocol serial number. When the rearrangement queue Q' is received or the scheduling time limit is reached, the network layer agent Informing the link layer agent to start the automatic retransmission request transmission with decreasing priority to the rearrangement queue Q';

The link layer agent further divides the rearrangement queue Q' into wireless link units with different priorities according to the two-step priority judgment rule, and performs tentative transmission to the wireless link units in order of decreasing priority;

Before the link layer agent tries to send a radio link unit, it will decide whether to send it according to the upper limit of the transmission times of the radio link unit.

The calculation method of described fault-tolerant importance measure is as follows: if can calculate GOP level transmission distortion according to sideband information, the importance level of the IP bag that GOP level transmission distortion is bigger so is higher; Otherwise according to default order, I frame type The IP packet of the IP frame type, the IP packet of the P frame type, and the IP packet of the B frame type correspond to three importance levels of high, medium, and low respectively, and the IP packet encoded earlier in the same type has a higher importance; the described The transmission distortion at the GOP level is the sum of the transmission distortion of the encoded frame and the transmission distortion of the uncoded frame caused by the IP packet loss and divided by the length of the IP packet.

The transmission distortion of the coded frame is the sum of the pixel-level transmission distortions of all pixels in the IP packet in the coded frame; the pixel-level transmission distortion is the product of pixel compensation error and error diffusion count; the pixel The compensation error is the absolute error between the reconstructed pixel value at the sending end and the reconstructed pixel value at the same position in the previous frame; the pixel error diffusion count is the total number of pixels directly or indirectly referring to the pixel in the encoded frame, by It is obtained by iteratively accumulating the number of pixels on the prediction link where the pixel is located in the frame.

The transmission distortion of the unencoded frame is that the IP packet loss is estimated by error diffusion in the unencoded frame, which is obtained by a piecewise linear attenuation model or an inter-frame macroblock proportional attenuation model; in the piecewise linear attenuation model, from The frame-level transmission distortion from the frame where the IP packet is located to the subsequent frame associated with motion compensation presents a linearly decreasing attenuation characteristic frame by frame over time, and the decreasing step size is obtained by linear fitting of the frame-level transmission distortion of the encoded frame after the IP packet; In the inter-frame macroblock proportional attenuation model, the frame-level transmission distortion from the frame where the IP packet is located to the subsequent motion compensation associated frame presents a proportionally decreasing attenuation characteristic over time, and its decreasing scale factor is obtained by counting the inter-frame of the latest encoded frame. The macroblock ratio is obtained; the motion compensation-associated frame is a video frame that has a direct or indirect motion compensation prediction relationship with the frame where the IP packet is located.

The two-step priority decision rule is used to assign a unique priority for each wireless link unit in the DFS, including the following two rules: Rule 1, the wireless link unit belonging to a higher importance level IP packet has a higher priority. High priority; Rule 2, wireless link units in the same IP packet have descending priority according to the coding order.

The upper limit of the number of transmissions is the maximum allowable number of transmissions of the wireless link unit; if the actual number of transmissions of the wireless link unit is less than the upper limit of the number of transmissions, then send the wireless link unit; otherwise, stop sending the current DFS The wireless link unit in turn prepares to send the next DFS wireless link unit.

Beneficial effect

Due to the adoption of the above technical solution, compared with the prior art, the present invention has the following advantages and positive effects: In this method, the main communication nodes of the online video stream can work in division and cooperate, and jointly serve as stream primitives of different communication levels A selective retransmission mechanism is provided to realize priority transmission at the radio link unit level. In addition, the fault-tolerance importance measure can be used to evaluate the fault-tolerance importance of different IP packets under various delay constraints online, and can adapt to non-stationary video content. Compared with the existing selective transmission scheme, the present invention can transfer the transmission distortion of a single video stream to a relatively unimportant part in each DFS, thereby protecting a relatively important part in the same DFS. This effectively improves the overall reconstruction quality of online video streams.

Description of drawings

Fig. 1 is the online video streaming system architecture of the present invention method and its main communication node schematic diagram;

Fig. 2 is a schematic diagram of the DFS structure of the present invention;

Fig. 3 is the functional block schematic diagram of the packet scheduling based on fault-tolerant importance of the present invention;

Fig. 4 is the schematic diagram of the GOP level transmission distortion adopted by the importance classification module of the present invention;

Fig. 5 is the work flowchart of the ARQ of decreasing priority among the present invention;

Fig. 6 is a schematic diagram of the application of the method of the present invention.

Detailed ways

Below in conjunction with specific embodiment, further illustrate the present invention. It should be understood that these examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the teachings of the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of the present application.

The invention relates to a cross-node controlled online video stream selective retransmission method, which can enhance the error control performance of the online video stream. The delay limit is divided into DFS one by one as the basic object of selective retransmission; the sender determines the importance level of different IP packets in DFS according to the fault tolerance importance measure, and through the packet scheduling based on the fault tolerance importance Provide selective retransmission of IP packets; further, the base station determines the priority of different wireless link units in DFS according to the two-step priority decision rule, and provides selective retransmission of wireless link units through the ARQ method with decreasing priority pass.

Fig. 1 is an online video streaming system architecture and its main communication nodes applicable to the method of the present invention, and the selective retransmission across nodes is jointly completed by two communication nodes, the sending end and the base station. The online video stream comes from an encoder that uses motion compensation prediction such as H.264/AVC or MPEG-4. The sending end performs packet scheduling based on the importance of error tolerance to provide selective retransmission of IP packets. On this basis, the base station performs ARQ with decreasing priority is used to provide selective retransmission of radio link units.

FIG. 2 is a schematic diagram of the structure of DFS in FIG. 1. DFS is a video frame segment divided for selective retransmission of online video streams, and serves as a basic object of cross-node selective retransmission. For the sake of brevity, the specific implementation manners of the video stream GOP in FIG. 2 , FIG. 4 and FIG. 6 do not include the B frame type. With the increase of cellular network bandwidth, the entire transmission delay of a single video stream is about hundreds of milliseconds or even lower. With the improvement of processor performance, the actual encoding time or decoding time of a video frame is lower than 33 milliseconds. A single video stream application usually allows a start-up delay of about a few seconds, while the video frame interval of online encoding is about 33 milliseconds to 100 milliseconds, so each DFS can contain several video frames according to the delay constraint. The delay constraint should be Less than the difference between the startup delay and the transmission delay. Under a given delay constraint, each DFS contains an equal number of video frames. At the sending end, each frame is encoded into several IP packets; at the base station, each IP packet is further encapsulated into several wireless link units. DFS and GOP are two different concepts. For the convenience of description, the length of DFS " _ND " ( _ND ≥ 1) indicates the number of frames in each DFS. In Figure 2, _ND is equal to 3; " _NG " indicates the number of frames in each GOP, and its typical value is 12 , 24, 15 and 30. The value of N _D can be greater than N _G , so that one DFS includes frames of multiple GOPs. Video frames and image slices are numbered according to the encoding order, "n" (n≥1) indicates the frame number incremented according to the encoding order, video frame n is divided into "M _n " (M _n ≥1) image slices, "P _n,m ”indicates the IP packet corresponding to the “m”th (1≤m≤M _n ) image slice of the nth video frame, and “S _n,m ”indicates the number of bytes in the P _n,m packet length. In the wireless access network, P _{n, m} is divided into several wireless link units with fixed length, “K _{n, m} ” indicates the number of wireless link units contained in P _{n, m} , and its value is equal to IP The ratio of packet length to radio link unit length.

(p≤n＜“p+N_D”)”。为了确定当前DFS中每个IP包的相对重要性，重要性归类模块需要计算P_n，m的GOP级传输失真。“GTD(n，m)”表示IP包P_n，m单独损失导致的GOP级传输失真，即从第n帧到该GOP最后一帧的传输失真之和。FIG. 3 is a schematic diagram of functional blocks of packet scheduling based on importance of fault tolerance in FIG. 1 . For online video streaming, the encoder and the sender are two loosely coupled communication nodes. The sending end uses the sideband information extraction module to obtain and store the sideband information. Among them, the side information includes packing method, error compensation method, DFS length and real-time encoding result, and these information will be provided to the importance classification module. Specifically, the packetization method follows the packetization rules of RFC 3894, the error compensation method adopts the previous frame copy method, and the DFS length _ND is determined according to the delay constraints allowed by the current application. The IP packets output by the encoder are first input to the frame set scheduling module, which divides a certain video stream into DFSs one by one according to the encoding order, and assembles the IP packets from the same DFS into an encoding queue according to the encoding order

(p≤n<"p+N _D ")". In order to determine the relative importance of each IP packet in the current DFS, the importance classification module needs to calculate P _{n, the GOP level transmission distortion of m} . "GTD(n, m)" indicates the GOP-level transmission distortion caused by the loss of IP packets P _{n and m} alone, that is, the sum of the transmission distortion from the nth frame to the last frame of the GOP.

表示i_n，m在发送端的重建像素值，可从编码器的编码环节中得到。绝对误差均值(简称“MAD”)是视频编码中常用的度量准则，其值越小，表示性能越好。若N_c小于N_g，表明GTD(n，m)包含已编码帧的传输失真和未编码帧的传输失真；否则，GTD(n，m)仅包含已编码帧的传输失真。具体地，重要性归类模块根据式(1)计算GTD(n，m)的值。Fig. 4 is a schematic diagram of GOP-level transmission distortion adopted by the importance classification module in Fig. 3. The loss of an IP packet will cause transmission errors to diffuse in the reconstructed video until the last frame of the GOP where the IP packet is located. As shown in Figure 4, the error caused by the black rectangular area in frame 2 can spread in the temporal and spatial directions of the reconstructed video until an I frame appears to eliminate this effect. For the convenience of description, “N _c ” indicates the frame number of the last encoded frame of the current DFS, “N _g ” indicates the frame number of the last frame of the GOP where P _{n, m} is located, and the GOP where P _{n, m} is located may be all encoded frame. The transmission distortion before frame N _c (including frame N _c ) needs to analyze the real-time encoding result provided by the sideband information extraction module, and the transmission distortion after frame N _c needs to be estimated by a certain prediction model. "GTD _e (n, m)" (n≤N _c ) indicates the transmission distortion of the encoded frame caused by the loss of IP packet P _{n, m} , and "GTD _f (n, m)"(n>N _c ) indicates that the IP packet Transmission distortion of uncoded frames due to P _n,m loss. "i _n,m " means a pixel in P _n,m ,

Indicates the reconstructed pixel value of i _{n, m} at the sending end, which can be obtained from the encoding link of the encoder. Mean Absolute Error ("MAD" for short) is a commonly used measurement criterion in video coding, and the smaller the value, the better the performance. If N _c is smaller than N _g , it means that GTD(n,m) includes both the transmission distortion of the encoded frame and the transmission distortion of the uncoded frame; otherwise, GTD(n,m) only includes the transmission distortion of the encoded frame. Specifically, the importance classification module calculates the value of GTD(n, m) according to formula (1).

$\mathrm{<span\; class="notranslate">\; GTD</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\frac{{\mathrm{<span\; class="notranslate">\; GTD</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; GTD</span>}}_{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}}& <span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>\\ \frac{{\mathrm{<span\; class="notranslate">\; GTD</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}}& <span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Further, GTD _e (n, m) is the sum of pixel-level transmission distortions of all pixels in P _{n, m} in the encoded frame. Pixel-level transmission distortion is the product of the pixel compensation error and the error diffusion count: the pixel compensation error is the absolute error of the reconstructed pixel value at the sending end and the reconstructed pixel value at the same position in the previous frame; The total number of pixels directly or indirectly referencing the pixel (including the pixel itself), that is, by tracking the motion vector information that finally points to the pixel in the encoded frame, iteratively accumulates the total number of pixels on all prediction links pointing to the pixel. Represents the compensation error of pixel i _{n, m} , "SRC _k (i _{n, m} )" represents the error diffusion count from frame n to frame k caused by the loss of i n _{, m} , which can be iteratively accumulated in the motion of i _{n, m} It is obtained by compensating the number of pixels on the prediction path, both of which belong to side information. Since pixel-level transmission distortion is accumulative within an IP packet, the importance classification module calculates the value of GTD _e (n, m) according to formula (2).

${\mathrm{<span\; class="notranslate">\; GTD</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; SRC</span>}}_{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

GTD _f (n,m) is the error diffusion distortion estimate of P _n,m loss in uncoded frames, which cannot be obtained by tracking motion vectors. The present invention provides two models to estimate the GTD _f (n, m) value respectively, that is, a piecewise linear attenuation model or an inter-frame macroblock proportional attenuation model, wherein the piecewise linear attenuation model is usually more accurate in estimation, but the complexity is higher.

In the piecewise linear attenuation model, the transmission distortion from the frame where P _{n, m} is located to the subsequent motion compensation associated frame presents a linearly decreasing attenuation characteristic frame by frame over time, and "λ(n, m)" means that the loss of P _{n, m} results in The decreasing step size of the frame-level transmission distortion of the uncoded frame, whose value is obtained by linear fitting of the error diffusion count of the coded frame by Equation (3).

$\mathrm{\&lambda;}(\mathrm{no},m)=\frac{\underset{i_{\mathrm{no},m}\&Element;P_{\mathrm{no},m}}{\mathrm{\&Sigma;}}|{\hat{i}}_{\mathrm{no},m}-{\hat{i}}_{\mathrm{no}-1,m}|\&Center\; Dot;{\mathrm{SRC}}_{\mathrm{no}}\left(i_{\mathrm{no},m}\right)-\underset{i_{\mathrm{no},m}\&Element;P_{\mathrm{no},m}}{\mathrm{\&Sigma;}}|{\hat{i}}_{\mathrm{no},m}-{\hat{i}}_{\mathrm{no}-1,m}|\&CenterDot;[{\mathrm{SRC}}_{N_c}\left(i_{\mathrm{no},m}\right)-{\mathrm{SRC}}_{N_c-1}\left(i_{\mathrm{no},m}\right)]}{N_c-\mathrm{no}}$ (n<N _c ) (3)

Based on the above-mentioned decreasing step size, the importance classification module can estimate the sum of transmission distortion from frame "N _c + 1" to frame N _g in the same GOP by formula (4):

${\mathrm{<span\; class="notranslate">\; GTD</span>}}_{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; \{</span>\underset{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; SRC</span>}}_{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; SRC</span>}}_{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>& <span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; )</span>\\ \underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; SRC</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>& <span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

In the above formula, considering the special case of "n=N _c ", "λ(l,m)"(l<n) represents the decreasing step size corresponding to the IP packet P l _{,m closest to P n ,m} .

In the inter-frame macroblock proportional attenuation model, the frame-level transmission distortion caused by P _n,m loss presents a frame-by-frame proportional attenuation characteristic in the subsequent motion-compensated associated frames of frame N _c , and its decreasing scale factor is calculated by statistically recently encoded The inter-macroblock ratio of the frame is obtained. In online video streams encoded with motion compensation prediction, all macroblocks can be classified into two categories: inter-frame macroblocks based on inter-frame motion estimation and intra-frame macroblocks without inter-frame motion estimation. "r(n)" represents the ratio of inter-frame macroblocks in frame n, that is, the ratio of the number of inter-frame macroblocks to the total number of macroblocks per frame. n ₁ and n ₂ (n ₂ <n ₁ <N _c ) represent the two coded P frames closest to frame N _c . Based on the above-mentioned inter-frame macroblock ratio, the importance classification module can also estimate the sum of transmission distortion from frame "N _c + 1" to frame N _g in the same GOP by formula (5):

${\mathrm{<span\; class="notranslate">\; GTD</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\underset{\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\underset{\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; \&Center\; Dot;</span>\underset{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; i</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

In the above formula, "β ₁ " and "β ₂ " (β ₁ +β ₂ =1) are two weighting parameters, and the initial values are 0.8 and 0.2 respectively. After the frame encoding is completed, _β1 and _β2 can be updated according to the actual inter-frame macroblock ratio through least square fitting.

(q≤n＜“q+N_D”)，其中重要性等级高的IP包在重排队列Q′的前面，P_q，1具有最高的重要性等级。在形成Q′之后，包重排模块立即将P_q，1和的序列号作为信令信息提前发送给基站。接着，串并转换模块将递增的实时传输协议(简称“RTP”)序列号依次分配给Q′中的所有IP包，并在同一调度时隙内按RTP序列号递增的顺序将这些IP包发送出去。在发送完毕后，重排队列Q′在发送端将被缓存一段时间以用于IP包的选择性重传。基于DFS结构，16位的RTP序列号可提供65536个重要性等级，这远高于目前H.264/SVC提供的最大64个重要性等级。In the same DFS, an IP packet with greater transmission distortion at the GOP level has a higher importance level. The importance classification module determines the relative importance of IP packets P _{n, m} in the DFS according to GTD(n, m), and assigns P _{n, m} a unique importance level "c(n, m)". Next, the packet rearrangement module sorts all the IP packets in the encoding queue Q in the order of decreasing importance levels, thus forming a new rearrangement queue

(q≤n<"q+N _D "), wherein the IP packets with high importance level are in front of the rearrangement queue Q', and P _q,1 has the highest importance level. Immediately after forming Q′, the packet rearrangement module combines P _{q, 1} and The serial number is sent to the base station in advance as signaling information. Next, the serial-to-parallel conversion module assigns increasing real-time transport protocol (referred to as "RTP") sequence numbers to all IP packets in Q' in turn, and sends these IP packets in the same scheduling time slot in the order of increasing RTP sequence numbers go out. After sending, the rearrangement queue Q' will be buffered for a period of time at the sending end for selective retransmission of IP packets. Based on the DFS structure, the 16-bit RTP sequence number can provide 65536 importance levels, which is much higher than the maximum 64 importance levels currently provided by H.264/SVC.

It should be noted that when the GOP-level transmission distortion cannot be calculated according to the sideband information, IP packets can be given a unique importance level according to the default order, that is, IP packets of I frame type, IP packets of P frame type, and B frame type IP packets. The IP packets of the frame type correspond to three importance levels of high, medium, and low respectively, and the IP packets encoded first in the same frame type have a higher importance level.

After the transmitting end provides selective retransmission with IP packet precision, the base station will further provide selective retransmission with radio link unit precision. The network layer agent of the base station is used to receive the IP packets of the rearrangement queue Q', and restore the sequence of the IP packets in the serial-to-parallel conversion module according to the RTP sequence number. Assume that the reordering queue Q' contains C _Q' IP packets, of which only A% of the IP packets are allowed to be retransmitted. Within the scheduling time limit of Q', the network layer agent will select [C _Q' × A%] IP packets with higher importance level to request retransmission, and send the corresponding Negative Acknowledgment ("NACK") The packet is sent to the sender, and the NACK packet adopts the RFC 4585 specification. When the rearrangement queue Q' is received or the scheduling time limit is up, the network layer agent immediately notifies the link layer agent to start executing the ARQ transmission mechanism with decreasing priority for the current DFS.

个无线链路单元，链路层代理根据两步优先级判决规则为这L_Q′个无线链路单元分配唯一的优先级。其中，两步优先级判决规则用于为DFS中每个无线链路单元分配唯一的优先级，包括以下两个规则：规则一，属于更高重要性等级IP包的无线链路单元具有更高的优先级；规则二，同一IP包内的无线链路单元按编码顺序具有递减的优先级。按照编码顺序，“l_n，m，k”表示IP包P_n，m的第“k”(1≤k≤K_n，m)个无线链路单元。从一个DFS中具有最高优先级的无线链路单元l_q，1，1开始，链路层代理按照优先级递减的顺序依次对当前DFS的无线链路单元进行尝试性的发送。Based on the data link layer technology of the wireless access network, the link layer agent of the base station further divides the IP packet P _{n, m} into K _{n, m} wireless link units with fixed length, so that the current DFS has

The link layer agent assigns a unique priority to these L _Q′ wireless link units according to the two-step priority decision rule. Among them, the two-step priority decision rule is used to assign a unique priority to each wireless link unit in DFS, including the following two rules: Rule 1, the wireless link unit belonging to a higher importance level IP packet has a higher priority; Rule 2, wireless link units in the same IP packet have descending priorities according to the coding order. According to the coding sequence, "l _n,m,k " indicates the "k"th (1≤k≤Kn _,m ) wireless link unit of the IP packet P _n,m . Starting from the radio link unit l _{q, 1, 1} with the highest priority in a DFS, the link layer agent makes tentative transmissions to the radio link units of the current DFS in descending order of priority.

FIG. 5 is a flow chart of the ARQ with decreasing priority in FIG. 1 . In the wireless access network, the round-trip delay of the data link frame is about tens of milliseconds. If the ARQ information of the current wireless link unit is not received after the round-trip delay, the link layer agent will immediately send the video stream The next wireless link unit. A certain wireless link unit may have a certain number of transmission opportunities, and the link layer agent needs to accumulate the actual transmission times of the wireless link unit. "A _{n, m, k} " (A _{n, m, k} ≥ 0) indicates the actual number of transmissions of the wireless link unit l _{n, m, k,} the initial value is 0, every time l _{n, m, k} is transmitted, A _{n, m, k} are incremented by 1. "Φ _{n, m, k} " (Φ _{n, m, k} ≥ 0) indicates the upper limit of transmission times of the wireless link unit l _{n, m, k,} which is sent by the link layer agent for the first time l _{n, m, k} It can be calculated according to formula (6).

表示不大于x的最大整数。在优先级递减的ARQ机制中，如果当前Φ_n，m，k等于0，链路层代理停止发送当前DFS的无线链路单元，也就是丢弃当前DFS中所有未传的无线链路单元，转而准备发送下一DFS的无线链路单元；如果当前Φ_n，m，k大于0，链路层代理将尝试发送l_n，m，k，若无法纠正的物理层错误导致l_n，m，k损失，接收端将立即反馈该无线链路单元的ARQ信息，链路层代理可对l_n，m，k进行最多“Φ_n，m，k-1”次的重传。如果无线链路单元l_n，m，k被成功接收，链路层代理将根据式(6)计算下一无线链路单元“l_n，m，k+1”的传输次数上限“Φ_n，m，k+1”。如果当前DFS的L_Q′个无线链路单元都被成功接收，基站的链路层代理将做好下一个DFS的传输的发送准备。如果较高优先级的无线链路单元用完了当前DFS所允许的最大传输次数，链路层代理不得不丢弃一些较低优先级的无线链路单元。In the above formula, the real number "η" (η≥0) represents the maximum average number of transmissions allowed by each radio link unit, which represents the total link layer transmission resources in the radio access network. symbol

Indicates the largest integer not greater than x. In the ARQ mechanism with decreasing priority, if the current Φ _{n, m, k} are equal to 0, the link layer agent stops sending the wireless link units of the current DFS, that is, discards all untransmitted wireless link units in the current DFS, and transfers to And the wireless link unit that is going to send the next DFS; if the current Φ _{n, m, k} is greater than 0, the link layer agent will try to send l _{n, m, k} , if the uncorrectable physical layer error causes l _{n, m, k} loss, the receiving end will immediately feed back the ARQ information of the wireless link unit, and the link layer agent can retransmit l _{n, m, k} at most "Φ _{n, m, k-1} " times. If the wireless link unit l _{n, m, k} is successfully received, the link layer agent will calculate the _upper limit "Φ _{n, m,k+1} ". If all the _LQ' radio link units of the current DFS are successfully received, the link layer agent of the base station will prepare for the transmission of the next DFS transmission. If higher priority RLUs use up the maximum number of transmissions allowed by the current DFS, the link layer agent has to discard some lower priority RLUs.

Finally, a simplified application embodiment of the present invention is introduced with reference to FIG. 6 to further illustrate the idea of the present invention. Each DFS contains 3 video frames (N _D =3) in turn, each video frame is encoded into 2 IP packets with similar length, and each IP packet is further encapsulated into 2 wireless link units. After the encoder completes the encoding of the third frame, the 6 IP packets in the first DFS will form an encoding queue Q={P _1,1 ; P _1,2 ; P _2,1 ; P _2,2 ; P _{3, 1} ; P _{3, 2} }. By comparing the respective GOP-level transmission distortions, each IP packet in the first DFS will be assigned a unique importance level, i.e. {c(1,1)=6; c(1,2)=5;c (2,1)=2; c(2,2)=1; c(3,1)=3; c(3,2)=4}. According to the descending order of GTD(n, m) packet by packet, the packet rearrangement module rearranges these 6 IP packets, thus forming a new rearrangement queue Q'={P _1,1 ; P _1,2 ; P _{3 , 2} ; P _{3, 1} ; P _{2, 1} ; P _{2, 2} }. Assume that the network layer agent of the base station has completely received the rearrangement queue Q', and n is equal to 1. Although the physical layer error period occurs during the transmission of radio link units l _1,1,2 , l _1,2,2 , l _3,2,1 , l _3,2,2 , the ARQ mechanism with decreasing priority will These radio link units are retransmitted on a best-effort basis according to the upper limit of their transmission times. After a higher priority radio link unit (i.e. l _1,1,2 , l _1,2,2 , l _3,2,1 , l _3,2,2 ) is successfully received, the link layer agent actively Lower priority radio link units (ie l _2,1,1 , l _2,1,2 , l _2,2,1 , l _2,2,2 ) are discarded to balance the limited total transmission resources. In this way, transmission errors are finally transferred to relatively unimportant parts in DFS, enabling selective retransmission of a single video stream in the case of non-stationary video content and burst channel errors.

Claims (8)

1. the Online Video stream selectivity repeating method of a cross-node control is characterized in that, may further comprise the steps:

(1) according to starting and the transmission delay restriction; Transmitting terminal is divided into the basic object that time-delay constraint frame set one by one retransmits as selectivity with the single video flowing of online coding; Wherein, The set of said time-delay constraint frame is to select to retransmit the frame of video segmentation that marks off for single video flowing, wherein, includes the frame of video of equal amount in each time-delay constraint frame set; At transmitting terminal, each frame of video is encoded into P IP bag, P >=1; In the base station, each IP encapsulates and further is packaged into K the Radio Link unit with regular length, K >=1;

(2) transmitting terminal is as first communication node, and the packet scheduling of at first carrying out based on fault-tolerant importance to provide the selectivity of IP bag to retransmit for the set of time-delay constraint frame;

(3) based on the importance rate of IP bag, the base station is as the second communication node, and the automatic repeat requests mode that execution priority is successively decreased to provide the selectivity of Radio Link unit to retransmit for the set of time-delay constraint frame.

2. the Online Video stream selectivity repeating method of cross-node control according to claim 1 is characterized in that the packet scheduling based on fault-tolerant importance in the described step (2) may further comprise the steps:

Obtain and store side information; Described side information comprises packing manner, mistake compensation way and time-delay constraint frame set length and instant coding result;

The IP bag that will come from the set of same time-delay constraint frame assembles a coding formation Q by coded sequence;

Estimate to confirm that according to fault-tolerant importance each IP wraps in the relative importance among the affiliated coding formation Q, and give each IP bag a unique importance rate;

Order according to importance rate successively decreases is resequenced to all the IP bags among the coding formation Q, obtains a new queue and rearrangement Q thus ^'

The real time transmission protocol serial that will increase progressively is successively decreased by importance rate and is distributed to the IP bag in the queue and rearrangement in proper order successively,

And the order that in same time slot scheduling, increases progressively by real time transmission protocol serial sends queue and rearrangement Q '.

3. the Online Video stream selectivity repeating method of cross-node control according to claim 2 is characterized in that the base station in the described step (3) comprises network layer agency and link layer agency; May further comprise the steps in the automatic repeat requests mode that described priority is successively decreased:

Receive queue and rearrangement Q ' through described network layer agency; And recover IP according to real time transmission protocol serial and wrap the order after resetting; When queue and rearrangement Q ' finishes receiving or dispatch the time limit when reaching, described network layer agent advertisement described link layer agency begin queue and rearrangement Q ^'Carry out the automatic repeat requests transmission that priority is successively decreased;

Link layer agency further is divided into the Radio Link unit with different priorities according to two step priority decision rules with queue and rearrangement Q ', and the order of successively decreasing according to priority is successively to the transmission of Radio Link unit trial property;

Whether the link layer agency will send according to the number of transmissions upper limit decision of this Radio Link unit before attempting sending a Radio Link unit.

4. the Online Video stream selectivity repeating method of cross-node control according to claim 2 is characterized in that the computational methods that described fault-tolerant importance is estimated are following:

If can computed image group level transmission distortion according to side information, the importance rate of the IP bag that image sets level transmission distortion is big more so be high more; Otherwise according to default order, the IP of I frame type wraps, the IP of P frame type wraps and the respectively corresponding high, medium and low three kinds of importance rates of IP bag of B frame type, and the IP that elder generation encodes in same type bag has higher importance; Described image sets level transmission distortion be the transmission distortion of the coded frame that causes of IP bag loss and not after the transmission distortion addition of coded frame divided by this IP packet length.

5. the Online Video stream selectivity repeating method of cross-node according to claim 4 control is characterized in that, the transmission distortion of described coded frame is the Pixel-level transmission distortion sum of all pixels in coded frame in the IP bag; Described Pixel-level transmission distortion is the product of pixel compensation mistake and error diffusion counting; Described pixel compensation mistake is the absolute error that transmitting terminal is rebuild the reconstruction pixel value of pixel value and former frame same position; Described pixel error diffusion counting be in coded frame directly or indirectly with reference to the total pixel number of this pixel, obtain through the pixel count that in coded frame, adds up iteratively on this pixel place prediction link.

6. the Online Video stream selectivity repeating method of cross-node control according to claim 4; It is characterized in that; The transmission distortion of described not coded frame is that IP bag is lost in not in the coded frame error diffusion and estimates, obtains through piecewise linearity attenuation model or inter macroblocks ratio attenuation model; In described piecewise linearity attenuation model; Frame presents the attenuation characteristic of successively decreasing by frame linearity in time to the frame level transmission distortion of subsequent motion compensation disassociation frame from IP bag place, its decrement step size through to after the IP bag the frame level transmission distortion of coded frame carry out linear fit and obtain; In described inter macroblocks ratio attenuation model; Frame presents the attenuation characteristic that ratio is successively decreased in time to the frame level transmission distortion of subsequent motion compensation disassociation frame from IP bag place, and the inter macroblocks ratio of coded frame obtains its scale factor that successively decreases through adding up recently; Described motion compensation disassociation frame is the frame of video that has direct or indirect motion compensated prediction relation with this IP bag place frame.

7. the Online Video stream selectivity repeating method of cross-node control according to claim 3; It is characterized in that; Described two step priority decision rules are used to the unique priority of each Radio Link unit distribution in the set of time-delay constraint frame; Comprise following two rules: rule one, the Radio Link unit that belongs to higher importance rate IP bag has higher priority; Rule two, the Radio Link unit in the same IP bag has the priority of successively decreasing by coded sequence.

8. the Online Video stream selectivity repeating method of cross-node control according to claim 3 is characterized in that the maximum that the described the number of transmissions upper limit is this Radio Link unit allows the number of transmissions; If the actual transmissions number of times of this Radio Link unit less than the described the number of transmissions upper limit, then sends this Radio Link unit; Otherwise, stop to send the Radio Link unit of current delay constraint frame set, then be ready for sending the Radio Link unit of next time-delay constraint frame set.

Images