CN-121984952-A - Self-adaptive forward error correction method for real-time audio and video stream transmission process

CN121984952ACN 121984952 ACN121984952 ACN 121984952ACN-121984952-A

Abstract

The invention provides a self-adaptive forward error correction method for a real-time audio and video streaming transmission process, which can dynamically adjust error correction redundancy configuration according to the change of network packet loss characteristics so as to improve the data transmission reliability in a complex network environment. Specifically, in the real-time transmission process, the sending end carries out long-term trend estimation and short-time burst estimation on network packet loss by adopting a Kalman filtering algorithm and an exponential weighted moving average algorithm respectively based on a packet loss rate measured value fed back by the receiving end, and a packet loss prediction model with double time scales is constructed. Meanwhile, a dynamic safety redundancy boundary is introduced by calculating the statistical fluctuation characteristic of the historical packet loss rate, and a hysteresis state control mechanism is combined to smooth the redundancy adjustment process, so that frequent fluctuation is avoided. The method can reduce redundancy overhead in complex and changeable network environments, improve error correction capability, reduce packet loss rate, is suitable for real-time audio and video streaming transmission under various network conditions, and has good practicability and robustness.

Inventors

CUI JINXI
HUANG YUHUA
ZENG QINGXI

Assignees

南京航空航天大学

Dates

Publication Date: 20260505
Application Date: 20260119

Claims (6)

1. A self-adaptive forward error correction method oriented to a real-time audio and video streaming transmission process is characterized by constructing a closed-loop control mechanism based on double-time-scale packet loss prediction and network fluctuation rate decision and dynamically adjusting the forward error correction redundancy rate according to the network state. The method comprises the following steps: The method comprises the steps of 1, constructing a double-time-scale packet loss model prediction, namely, receiving a real-time packet loss rate feedback value fed back from a receiving end by a sending end, carrying out smooth processing on the packet loss rate through a Kalman filtering algorithm to obtain a trend packet loss rate representing a long-term stable state of a network, and carrying out weighted calculation on the packet loss rate through an exponential weighted moving average algorithm to obtain a burst packet loss rate representing a short-time burst change of the network, thereby realizing multi-time-scale perception of network packet loss behavior. And 2, constructing a dynamic marginal decision, namely calculating a fluctuation rate index of the network packet loss rate by adopting a recursive manner by a transmitting end based on a historical packet loss rate measured value fed back by a receiving end, and representing the discrete degree of the packet loss rate along with the change of time. And (2) determining a dynamic safety coefficient according to the fluctuation rate index, taking the product of the dynamic safety coefficient and the fluctuation rate as a safety margin, superposing the safety margin on the packet loss rate in the step (1), synthesizing a target effective packet loss rate, and further calculating to obtain the redundancy rate for forward error correction coding. And 3, constructing a holding control mechanism comprising a steady state and a burst state, triggering the holding mechanism when the burst packet loss characteristic is detected, locking the current higher redundancy level, carrying out limited attenuation or advanced release on the redundancy rate according to continuous low packet loss feedback during the holding mechanism, carrying out amplitude limiting control on the descending amplitude of the redundancy rate in a non-burst state so as to inhibit severe oscillation of the redundancy rate, and finally outputting a smooth and stable forward error correction redundancy rate for real-time audio/video stream coding.
2. The dual time scale packet loss model according to claim 1, wherein when constructing the dual time scale packet loss model prediction, the transmitting end calculates a trend packet loss rate p trend and a burst packet loss rate p brust according to a kalman filtering algorithm and an exponential weighted moving average algorithm respectively based on the same packet loss rate measurement value p t fed back by the receiving end, and determines a prediction reference packet loss rate p base according to the following rule: Wherein, alpha is a preset weighting coefficient and satisfies 0.5< alpha <1.
3. The dynamic marginal decision of claim 1, wherein the network packet loss fluctuation rate sigma t is recursively calculated based on a packet loss measurement value p t fed back by a receiving end, the packet loss fluctuation rate is used for representing the variation amplitude of the packet loss rate in a time dimension, the dynamic safety coefficient k t corresponding to the packet loss fluctuation rate sigma t is determined according to a preset interval where the packet loss fluctuation rate sigma t is located, and the packet loss safety margin is determined according to the following relation: m t ＝k t ·σ t And superposing the packet loss safety margin to the predicted packet loss rate p base obtained in the step 1 to obtain a target effective packet loss rate p eff : p eff ＝p base +m t The target effective packet loss rate is used for calculating the redundancy rate of the forward error correction coding.
4. The preset interval of the packet loss ripple σ t according to claim 3, wherein the dynamic security coefficient k t is determined according to a size segment of the network packet loss ripple σ t , specifically, the dynamic security coefficient k t takes a first preset value when the packet loss ripple σ t is smaller than a first threshold, the dynamic security coefficient k t takes a second preset value when the packet loss ripple σ t is greater than or equal to the first threshold and smaller than a second threshold, and the dynamic security coefficient k t takes a third preset value when the packet loss ripple σ t is greater than or equal to the second threshold, wherein the first preset value is greater than the second preset value and the second preset value is greater than the third preset value.
5. The steady-state or burst state of claim 1, wherein the network is distinguished to be in a steady state or a burst state according to the magnitude relation between the burst packet loss rate p brust and the trend packet loss rate, trend obtained in the step 1.
6. The hold control mechanism of claim 1, wherein the hold mechanism is triggered to perform a fade control on the redundancy rate when the system is to reduce redundancy and is in a packet loss burst state, and wherein the fast reduction control is performed on the redundancy rate when the system is in a plateau state.

Description

Self-adaptive forward error correction method for real-time audio and video stream transmission process Technical Field The invention aims to solve the problems of packet loss recovery and bandwidth utilization of real-time audio and video in an unstable network environment by adaptively adjusting forward error correction redundancy in the real-time audio and video streaming transmission process so as to cope with the transmission scene of the unstable complex network, and belongs to the technical fields of network communication and multimedia transmission. Background With the rapid development of applications such as real-time audio and video communication, cloud games, teleconferences and the like, the requirements of real-time media transmission based on a packet switched network on low time delay and high reliability are increasingly increased. However, in wireless network, mobile network and public network environments, link packet loss has obvious randomness and burstiness, and the conventional error correction mode relying on retransmission mechanism is difficult to meet the real-time requirement. Forward error correction (Forward Error Correction, FEC) is widely used in real-time audio and video transmission as a technique for recovering lost data without retransmission. By introducing a certain proportion of redundant data into the transmitting end, the receiving end can recover the original data by utilizing the redundant data when packet loss occurs, thereby reducing play jamming and picture distortion. However, existing forward error correction schemes mostly employ fixed redundancy rates or simple adaptive strategies based on a single packet loss indicator. On one hand, the fixed redundancy rate is difficult to adapt to the dynamic change of the network state, bandwidth waste is caused when the network is good, and enough protection cannot be provided when the network is deteriorated, on the other hand, the regulation method based on the instantaneous packet loss rate or the average packet loss rate is difficult to distinguish the long-term trend packet loss from the short-term burst packet loss, frequent oscillation or excessive amplification of the redundancy rate is easy to cause, and the system stability is further influenced. Disclosure of Invention The invention aims to provide a method for correcting the forward error in the prior art, which is applied to the traditional multimedia transmission field to a certain extent, but in the complex unstable network environment, the prior error correction method mostly adopts a fixed redundancy or a single packet loss index-based regulation mode, so that the method is difficult to simultaneously adapt to the transmission scene of the complex and unstable network, and the stable regulation of the redundancy rate in the real-time audio/video stream transmission process is still to be further studied. According to the self-adaptive forward error correction method provided by the invention, the long-term packet loss trend and the burst packet loss characteristic of the network are distinguished by constructing the double-time-scale packet loss prediction model, the traditional error correction adjustment method which only depends on the average packet loss rate or the instantaneous packet loss rate is difficult to realize fine control, and the method can realize effective adaptation to different network states. The invention further introduces a dynamic marginal decision and hysteresis maintenance state control mechanism based on the network fluctuation rate, and maintains the stability of redundancy adjustment when the network state fluctuates, thereby reducing the waste of bandwidth resources while ensuring the error correction performance. The method comprises the following steps of constructing a double-time scale packet loss model by using a Kalman filtering algorithm and an exponential weighted moving average algorithm, respectively predicting long-term trend and short-time burst characteristics of network packet loss, introducing a steady state and burst state, introducing a dynamic safety marginal decision mechanism based on the network packet loss fluctuation rate, finally generating a target effective packet loss rate, calculating redundancy rate required by forward error correction according to the target effective packet loss rate, and simultaneously smoothing and restraining an adjustment process of the redundancy rate by a hysteresis maintenance control mechanism, thereby effectively reducing redundancy cost while guaranteeing error correction performance. Further, in the dual time scale packet loss prediction process of the present invention, the sending end performs long-term trend estimation p trend on the packet loss rate by using a kalman filtering algorithm based on the packet loss rate measurement value fed back by the receiving end, and performs short-term burst estimation p brust on the packet loss rate by using an exponential w