Akce ČVUT

T. Matsumoto: Turbo Equalization to Lossless and Lossy Distributed Multiterminal Source Coding …

29.10.2019   14:30-16:00

A goal of this talk is to provide the audience with the knowledge about the relationship between relay systems and the Lossless/LossyDistributed Coding techniques for correlated sources. To achieve this goal, this lecture is started by the re-enforcement of understanding of turbo principle, especially frequency-domain soft cancellation minimum mean square error (FD SC-MMSE) based multiple-input multiple-output (MIMO) turbo equalization. This lecture uses a lot of extrinsic information transfer (EXIT) analysis to reveal the convergence properties of the FD SC-MMSE MIMO equalization, and identifies the optimal, close capacity achieving structure. It is shown that even with very simple serially concatenated convolution code with the component codes being very simple memory one codes can achieve near-capacity performance. Furthermore, the inner code, which is a very simple memory one recursive code, can eliminate the error floor due to the intersection of the EXIT curves, resulting in very sharp shape of the turbo cliff. This lecture also makes comparison of the shape of the EXIT curves with recursive and non-recursive convolutional codes.
This lecture then “add” intentionally binary errors randomly between the MIMO antennas, and analyzes the impacts of the “artificial errors”. It is shown that if the FD SC MMSE equalizer can utilize the error probability to modify the log likelihood ratio (LLR) in the vertical iteration, we can eliminate the effect of the “artificial errors”. The “artificial error” probability can be estimated only at the decoder side (no side information needed at the decoder side).
Obviously the “artificial errors” inserted in the connections between the antennas correspond to the “intra-link” errors in distributed lossy forwarding cooperative networks. Therefore, the last part of this lecture focuses on the introduction of Lossy Forward Relaying: in orthogonal and non-orthogonal transmission scenarios. Since the knowledge about the bit error probability of the source-relay node can be used as the correlation between the two frames, one from the source, and the other from the relay, we can well exploit the Slepian-Wolf theorem; We further expand conceptual bases of the lossless (Slepian Wolf) multi-terminal network design, and provides performance bounds theoretically as well as basic ideas for signal detection algorithms based on the turbo principle. This technique is referred to as Lossy Forward (LF) Relaying. This talk provides analytical techniques for the outage probability of the LF technique in fading channels by using the distributed Lossless Muli-terminal Source Coding with a helper. We also analyze the outage probabilities for various network topologies utilizing the LF techniques.
This talk further expands the LF techniques to Lossy case where a certain level of End-to-End distortion is accepted. We view such system with Berger-Tung’s Lossy Distributed Multi-terminal Source Coding with Helpers, where the outage analysis for the Chief Executive Officer (CEO) problem setup in fading channels is also introduced. Since solving those problems requires is highly mathematical background knowledge, this part will only be explained at the logic level, and mathematical details are explained, if needed, in the Q&A session.

Místo konání

T2:C3-434

Pořadatel

prof. Sýkora, K13137

Kontaktní osoba

prof. Sýkora, K13137, sykoraja@fel.cvut.cz, ‭+420 224 355 967‬