Deep Generative Model based Channel Agnostic Communication System for Efficient Data Transmission

In this modern era, where there is a great shift in momentum towards AI and data science, data-driven approach has substantiated to have diverse talents. This type of approach specifically in the field of wireless communication has proven to be highly efficient in modelling an end to end wireless communication channel with a completely unknown channel state information (CSI). Though there has been some work done in this direction they are not focusing on high dimension data like audio and image. Further with the tremendous growth in multimedia technologies and with higher resolution data being transmitted, there is an increasing need for bandwidth. Hence in this paper, we propose a novel idea where a Conditional Generative Adversarial Network is used to represent the channel effects of an end to end wireless communication channel which adapts to different noise levels with having the memory of the channel state information (CSI). Further to address this issue with bandwidth, Variational Adversarial Network is used where the input data is encoded into a condensed latent space and impacted with noise which is then passed through a Generative Adversarial Network which would be able to extract the original information. This reduces a lot of bandwidth and thereby increasing the overall data rate. To strengthen our proposal, we provide a comparative analysis with our conditional GAN and with extensive experiment results with both quantitative and qualitative analysis on images and voice signals.