Direkt zum Inhalt | Direkt zur Navigation

Kolloquium Informationstechnik

Kolloquium Informationstechnik WT2012 - Aktuelle Vortragstermine
Kolloquium Informationstechnik

Gebäude 35, Raum 1397

 

 

 

                    Programm als PDF Download

 

 

 

 

 

Kolloquium Informationstechnik
WT 2012

Termin: Montag 17:00 (falls nicht anders angegeben)
Ort: Universität der Bundeswehr, Gebäude 35, Raum 1397

 

16.01.2012     Prof. Dr. Tadashi Matsumoto
Japan Institute of Advanced Science and Technologies (JAIST), Nomi, Ishikawa, Japan und University of Oulu, Finland
An IDMA Technique using Near-Capacity Achieving BICM-ID with Extended Mapping and Irregular Repetition Code
   
 6.02.2012 Dr. Nobuhiko Miki
NTT DOCOMO Research Labs, Yokosuka Research Park, Japan
Investigation on picocell deployments on heterogeneous network
   
19.03.2012 Prof. Dr.-Ing. Robert Fischer
Universität Ulm
Entzerrungsverfahren für Kanäle mit mehreren Ein- und Ausgängen: der Weg von der Gram–Schmidt–Orthogonalisierung über Sortierung und Quantisierung hin zu Lattice–Reduktion
   

 

Prof. Dr. Tadashi Matsumoto
An IDMA Technique using Near-Capacity Achieving BICM-ID with Extended Mapping and Irregular Repetition Code

Abstract:

Bit-Interleaved Coded Modulation with Iterative Detection/Decoding (BICM-ID) has been recognized as being a bandwidth efficient coded modulation and transmission scheme. D.Zhao et al. proposed a yet very simple, close Shannon-limit achieving BICM-ID system. It uses very simple codes, irregular repetition and single parity check codes, combined with Extended Mapping (EM). With this structure, turbo cliff happens at a value range of Signal-to-Noise power Ratio (SNR), 1 s 2dB to the Shannon limit. Even having known that the key role played towards the optimal design of the proposed code is the degree allocation for variable nodes, in the original technique described above, though, the irregular allocation to the node degrees were determined only empirically. Another fundamental drawback of the original technique is that it still suffers from error floor. This is simply because the demapper EXtrinsic Information Transfer (EXIT) curve does not reach the top-right (1,1) Mutual Information (MI) point. This talk shows that the problem of the optimal degree allocation for the proposed BICM-ID technique can be solved by using Linear Programming (LP). Furthermore, this talk introduces the use of partial accumulator, combined with the original system. With this technique, we can lift up the right-most point of the demapper curve to the (1.1) MI point, and thereby the convergence tunnel opens until (1,1) MI point. Based on those two optimization techniques, this talk further investigates a technique to identify optimal labeling pattern for EM, so that we can flexibly change the shape of the demapper EXIT curve. This talk introduces EXIT-constraint Binary Switching Algorithm (EBSA) and unbalanced labeling. It is shown through EXIT analysis that using EBSA and unbalanced labeling, demapper and decoder EXIT curves closely match, and thereby, it can achieve near-capacity performance. Simulation results using proposed techniques are also given in this talk. Using the proposed optimization techniques, threshold SNR only 0.5dB away from the Shannon limit even though its required complexity is very low; the complexity is at the same level as a turbo code with only memory-2 constituency codes.

Since, as noted above, EBSA jointly optimize the code and mapping rule.  We apply this technique to IDMA where entire available bandwidth expansion factor is used for coding, hence no spreading is used.   The design of such "reverse L-shaped EXIT matching" (because of very low operation point in SNR) is made possible by EBSA, where  the demapper and the code EXIT curves are exactly matched, and BER threshold happens at the designed threshold.  Furthermore, we applied it to soft interference cancellation case.  In this case, EBSA optimization was performed assuming all the multiuser interference is canceled, with which global soft cancellation and internal each-user's iterations are combined.  The system works very nicely, and the users can achieve very close performance to the multiple rate region bound.  At the final stage, the presentation makes technical comparison with the publication:
K. Kusume, G. Bauch, W. Utschick, “IDMA Vs. CDMA: Detectors, Performance and Complexity”, Global Telecommunications Conference, 2009. GLOBECOM 2009,
which uses combination of spreading and error correction code, and soft MMSE is used before error correction.  On the contrary, in our design, all the bandwidth is allocated to BICM-ID, and no MMSE needed.  Since our designed BICM-ID using EBSA is very powerful, exact matching is still possible for very low rate code (exhibiting "reverse L EXIT shape", and hence MMSE is not needed. Henceforth, our technique is much better in terms of power efficiency (lower SNR threshold) and spectrum efficiency (it can accommodate more users given the SNR operation point).

Keywords: BICM-ID, Single Parity Check Code, Irregular Repetition Code, Extended Mapping, Optimization, Partial Accumulator, EXIT based Binary Switching Algorithm, Interleaver Division Multiple Access.
 

 

Biography:

Harald Rohde received his Ph.D. in experimental quantum physics at the University of Innsbruck, Austria in 2001. Since then he worked for Siemens ICN, Siemens CT and lately for Nokia Siemens Networks, all in Munich, Germany. He worked in multiple fields on optical communication technologies, ranging from long haul technologies to in house communications. He authored or co-authored some 60 publications and has some 60 patents pending or granted in the field of optical communication. Currently his main research focus lies in the field of coherent optical access systems.

 

 

Dr. Nobuhiko Miki, NTT DOCOMO
Investigation on picocell deployments on heterogeneous network
 

Abstract:

Heterogeneous network, where different types of nodes such as pico eNodeBs, femto eNodeBs, and relay nodes are overlaid onto nacro eNodeBs using one radio interface based on LTE-Advanced, is now paid attention to improve the system throughput further. Techniques for pico eNodeBs deployments and its performance for Release 10/11 will be presented. Furthermore, this presentation also touches the techniques for future radio access towards future radio access beyond Rel-12.

 

Biography:

Nobuhiko Miki received his B.E. and M.E. degrees from Kyoto University, Kyoto, Japan in 1996 and 1998, respectively, and received his Dr. Eng. degree from Keio University, Yokohama, Japan in 2009. In 1998, he joined NTT Mobile Communications Network, Inc. (now NTT DOCOMO, INC.) His research interests include mobile communication systems.

 

 

Prof. Dr.-Ing. Robert Fischer, Universität Ulm
Entzerrungsverfahren für Kanäle mit mehreren Ein- und Ausgängen: der Weg von der Gram–Schmidt–Orthogonalisierung über Sortierung und Quantisierung hin zu Lattice–Reduktion
 

Abstract:

Seitdem in den 90er Jahren das Interesse an Kanälen mit mehreren Ein- und Ausgängen (multiple-input/multiple-output (MIMO)-Kanäle) stark gestiegen ist wurden in der Literatur ein Vielzahl von Entzerrungsverfahren für solche Kanäle untersucht. Insbesondere wurden Techniken, die von Kanälen mit Intersymbolinterferenzen bekannt waren, auf die räumliche Dimension übertragen, wie z.B. lineare Entzerrung, Entscheidungsrückgekoppelte Entzerrung (Decision-Feedback Equalization (DFE), auch als Successive Interference Cancellation (SIC) bekannt und das Grundprinzip hinter dem BLAST-System) und Maximum-Likelihood-Detektion.

Neben diesen klassischen Prinzipien sind neuere Ansätze basierend auf Lattice-Reduktion von besonderem Interesse. Diese Verfahren ermöglichen die optimale Diversitätsordnung bei nur sehr geringer Komplexität.

Im Vortrag wird die Verbindung des Gram-Schmidt-Verfahrens - aus der linearen Algebra zur Berechnung einer orthogonalen / orthonormal Basis eines Vektorraum bekannt - zu Entscheidungsrückgekoppelter Entzerrung und zu Lattice-Reduzierter Entzerrung beleuchtet. Es wird gezeigt, dass die Operationen "Quantisierung" und "Sortierung" eine wichtige Rolle spielen. Die Weiterentwicklung des universellen Werkzeugs "Gram-Schmidt-Verfahren" hin zu Lattice-Reduktion, im Speziellen zu der Variante des "LLL algorithms with deep insertions", wird erläutert. Der erreichbare Austausch zwischen Leistungsfähigkeit und Komplexität der Verfahren wird diskutiert.