Distortion Exponent in MIMO Channels with Feedback

TL;DR

This paper analyzes the distortion exponent for transmitting Gaussian sources over MIMO channels with feedback, revealing optimal strategies involving successive refinement coding and dynamic power allocation based on channel state.

Contribution

It characterizes the optimal exponential decay of distortion in MIMO channels with feedback and proposes a dynamic transmission strategy using successive refinement coding.

Findings

Optimal distortion exponent derived for all source-channel bandwidth ratios.

Successive refinement source coding combined with dynamic channel resource allocation is optimal.

Feedback enables adaptive transmission strategies that improve distortion performance.

Abstract

The transmission of a Gaussian source over a block-fading multiple antenna channel in the presence of a feedback link is considered. The feedback link is assumed to be an error and delay free link of capacity 1 bit per channel use. Under the short-term power constraint, the optimal exponential behavior of the end-to-end average distortion is characterized for all source-channel bandwidth ratios. It is shown that the optimal transmission strategy is successive refinement source coding followed by progressive transmission over the channel, in which the channel block is allocated dynamically among the layers based on the channel state using the feedback link as an instantaneous automatic repeat request (ARQ) signal.