Real-time video transmission adaptive forward error correction method and system based on deep learning

Abstract

The invention discloses a real-time video transmission adaptive forward error correction method and system based on deep learning, which can predict the network condition change trend in a future period of time when the network condition fluctuates and adaptively change the parameters of a forward error correction algorithm so as to ensure the transmission quality of real-time video data. The method mainly contributes to (1) learning a change rule of a network condition through a neural network model so as to predict a future network packet loss condition for a past network condition; and (2)adding a counter module to the model, and converting model output from a network feature sequence into a network packet loss rate, so that the model output is simplified, and the prediction accuracy is improved; and (3) setting an interval between past and future time periods during neural network learning and prediction, so that the problem of real-time feedback of network conditions is solved, and the neural network model can be used for real-time prediction in a real-time video transmission system.

Description

technical field [0001] The invention belongs to the technical field of network streaming media transmission, and in particular relates to a real-time video transmission method and system for realizing network self-adaptive forward error correction so as to reduce transmission packet loss. Background technique [0002] Live video streaming has become increasingly common in recent years. Online video streaming is expected to account for 82% of internet traffic by 2022. At the same time, more and more online videos will be in the form of live streaming, and real-time video communication is attracting more and more attention from users and researchers. However, packet loss is a critical issue in real-time video communication, as it causes distortion and decoding errors, thereby degrading the user's quality of experience. The International Telecommunication Union (ITU) standard limits the one-way delay to less than 200ms for real-time communication applications such as video co...

AI Technical Summary

Problems solved by technology

However, they all simply take the characteristics of the historical network directly as the prediction of the characteristics of the future network, ignoring the complex relationship that may exist between the history and the future.

When network conditions change frequently, these methods cannot achieve the expected effect of protecting data

Images