In the dynamic network environment, real-time video transmission based on datagram protocols will encounter a large number of packet loss issues. This paper proposed an adaptive forward error correction code method based on GA-BP neural network. This method recovered lost data packets by designing adaptive forward error correction codes. Firstly, the paper designed a frame level forward error correction framework, which mainly included video codec, RS forward error correction codec module, and forward error correction code redundancy calculation module, to simulate the general video transmission environment. Then, this paper designed the GA-BP model and applied it to the redundancy calculation of RS-FEC to achieve a frame level unequal protection scheme. Finally, based on GE channel simulation network, it generated packet loss data, trained offline model, and conducted experimental verification. The experiments results show that compared to StaticRS and DeepRS, this method can recover lost data packets more efficiently with lower redundancy, resulting in higher quality video.

[2023-07-05] Accepted Paper /doc/2023.04.0168/v1

[2023-12-05] Printed Article /doc/2023.04.0168/v2