Video-aware MIMO precoding with packet prioritization and unequal modulation

Video streaming over wireless networks can achieve large quality and capacity gains from transmission techniques that are aware of the video content. This paper proposes modifying conventional multiple-input multiple-output (MIMO) systems to realize these gains by introducing video-aware PHY layer decisions. Specifically, a new PHY layer packet prioritization method is introduced that splits video data among spatial streams based on the packet loss visibility and per-stream SNRs. The objective is to maximize throughput weighted by packet loss visibility, a metric coined perceived throughput. The globally-optimal splitting policy among streams is derived. Furthermore, unequal modulation per stream is proposed and visibility-based packet dropping is optimized to satisfy delay constraints. We derive the gains vs. conventional MIMO precoding and prove that the gain is the throughput averaged over streams divided by the throughput of the worst stream.