A comprehensive theoretical evaluation of the end-to-end performance of SoftCast-based linear video delivery schemes

Abstract

SoftCast-based linear video delivery (LVD) schemes have been proposed as an alternative to traditional video transmission schemes in wireless error-prone environments. The end-to-end performance of SoftCast-based schemes have been evaluated in Xiong et al. (2016), where a theoretical model based on the Peak Signal-to-Noise Ratio (PSNR) metric has been proposed. The latter is limited to the use of a Zero-Forcing (ZF) estimator at the receiver side, and does not consider bandwidth restrictions. Nevertheless, bandwidth restrictions are common and necessary in practice, especially when considering the transmission of video content. It is mandatory to take this aspect into consideration as it may drastically influence the received video quality. In this paper, we provide valid and significant extensions of the initial model. In total, three models are introduced taking into account both (1) bandwidth constraints (i.e., data compression applied), (2) the use of a Linear Least Square Error (LLSE) estimator instead of the ZF one as well as (3) the use of the optimal power allocation. We show that regardless of the bandwidth reduction applied, the type of estimator as well as the power allocation used, the end-to-end video quality can be accurately modeled and predicted at the transmitter according to the video content characteristics, the type of estimator used at the receiver and the channel conditions. The validity of these three models is assessed through extensive end-to-end simulations. These new models give solid theoretical guidelines for optimizing and studying the performance of linear video delivery schemes.