Content-Based Rate-Adaptive transfer of SVC-encoded video over MIMO communication systems

Abstract

A novel COntent-Based Rate-Adaptation (COBRA) model is proposed to improve the quality of video transmission over multiple-input multiple-output (MIMO) system with channel state information feedback (CSI). The model captures relationships among the video content type, video sub-stream bit-rates, achievable peak signal-to-noise ratios (PSNR), and channel bit error rate (BER). COBRA facilitates joint selection of physical layer parameters and video sub-stream configuration. Unlike existing methods, the COBRA model offers an efficient way of balancing losses in PSNR due to channel errors and non-transmission of some enhancements layers in a sub-stream, with a judicious mechanism of selecting sub-stream configurations and communication parameters. Simulations show that COBRA maintains video quality at acceptable levels over wide range of low channel signal-to-noise ratio (SNR) values. It achieves the same video PSNR as obtained by the complete stream transmission at 10 dB higher channel SNR.