Neural Network-Based optimization of Progressive Image Transmission in MIMO Systems

This paper considers unmanned aerial vehicles (e.g., drones and quadcopters) that accomplish missions of transmitting a large number of progressive images over air-to-ground multiple-input multiple-output (MIMO) links for surveillance applications, and special missions of public safety or emergencies. For the transmission of progressive images, the joint optimization of source and channel coding of a series of numerous packets has been a challenging problem. Further, the problem is more complicated if the space-time coding is also involved with the optimization in a MIMO system. This is because the number of ways of jointly assigning channel codes and space-time codes to progressive packets is much larger than that of solely assigning channel codes to the packets. Recently, Chang et al. applied a parametric approach to address such a problem, and proposed an optimization algorithm that exponentially reduces the computational complexities of the conventional exhaustive search. For resource-constrained aerial equipments, complexity is of particular importance due to hardware power consumption and size issues. To address such issues, we propose a neural network-based optimization method to further reduce computational complexities. We demonstrate that the nonlinear distortion-rate characteristics of the images, which are combined with wireless channel fading effects, can be analyzed and learned by a neural network. It is shown that compared to the algorithm proposed by Chang et al., our approach significantly reduces the computational complexities, while offering nearly identical peak-signal-to-noise ratio (PSNR) performances.