Source-channel separation for two-way interactive communication with fidelity criteria

TL;DR

This paper proves that for a two-way interactive communication system with dependent sources and separate channels, using source-channel separation architectures is optimal for achieving specified distortion levels.

Contribution

It establishes the optimality of source-channel separation in a two-way interactive setting with dependent sources and independent channels.

Findings

Separation-based architectures are optimal for the described interactive communication problem.

The problem involves dependent sources and independent DMC channels.

Optimal coding strategies can be designed based on this separation principle.

Abstract

Consider the channel coding problem where two users are interacting in order to communicate an i.i.d. source X1 from User 1 to User 2 with distortion D1 and an i.i.d. source X2 from User 2 to User 1 with distortion D2. X1 and X2 may be dependent. Communication occurs from User 1 to User 2 via a DMC C1 and from User 2 to User 1 over a DMC C2, where C1 and C2 act independently of each other. Communication occurs during each time slot between both users and each user can make a coding and decoding based on all past available knowledge. This interactive communication problem is formulated and it is proved that source-channel separation based architectures are optimal.