物联网中面向视频传输的GAN增强型语义通信系统OA

High-reliability video transmission in internet of things environments is crucial for critical applications such as intelligent surveillance and remote operations.However,the coupling of physical channel noise and visual semantic noise often affects video transmission quality.To effectively improve video transmission quality in internet of things environments while ensuring transmission reliability,a video semantic communication system robust to multi-dimensional noise based on Generative Adversarial Network(GAN)is proposed.The system integrates GAN into Deep Video Semantic Communication(DVSC)systems,incorporating Residual Channel Attention Block(RCAB)and multi-scale convolution-deconvolution structures to enhance semantic preservation and detail reconstruction capabilities under complex noise interference.Furthermore,it employs multi-scale discriminators and spectral normalization techniques to effectively stabilize adversarial training,mitigate mode collapse,and enhance the discriminator's capability in distinguishing video frames under adverse channel conditions.Finally,a multi-objective optimization function is designed by fusing pixel-level fidelity,perceptual similarity,and adversarial loss to collaboratively optimize video reconstruction quality.Experimental results demonstrate that compared to traditional methods,the proposed system improves video reconstruction quality and robustness under multi-dimensional noise conditions such as low Signal to Noise Ratios(SNR),with particularly significant quality improvements in high-noise environments.The effectiveness of the system in enhancing internet of things video communication reliability is validated,providing a novel solution for Ultra-Reliable and Low Latency Communications(URLLC).