Séminaire ASTRE/ICI : Chris Winstead
Titre du séminaire et orateur
Noise enhanced bit-flip decoding for LDPC codes on the Gaussian channel.
Chris Winstead, Utah State University (Dept. of Electrical and Computer Engineering), USA.
Date et lieu
Vendredi 15 mai 2015, 10h.
ENSEA, salle 384.
Recent results are presented for a highly efficient error correction algorithm known as Noisy Gradient Descent Bit Flipping (NGDBF). This algorithm improves on the previously known GDBF decoding method by injecting artificial "noise" into the operations, in effect "fighting noise with noise."
Our work is very similar to research on Probabilistic-GDBF (PGDBF) algorithm done at ETIS Lab, and some discussion will be given on the relationship between these algorithms.
Our new algorithm is applicable to Low-Density Parity Check (LDPC) codes, which are a key part of many high-performance communication standards. LDPC decoders typically employ advanced signal processing algorithms that are able to approach the theoretical Shannon limit in performance. These advanced algorithms, however, require resource-intensive implementations that lead to high power consumption and large silicon area, making them unsuitable for some applications. An alternative family of algorithms, collectively termed "bit-flipping" methods, perform decoding by exchanging single-bit messages, which gives them very low complexity. Although bit-flipping algorithms have been known for some time, they have had limited application due to their comparatively poor performance. Our latest results show that NGDBF can come very close to the best commercial algorithms on important standards, such as IEEE 802.3 (10GBase-T).
A tutorial introduction to the algorithm will be presented, along with a plausible implementation architecture, an analysis of the algorithm's theoretical characteristics, and a discussion of open problems.
Chris Winstead received the B.S. degree in Electrical and Computer Engineering from the University of Utah in 2000, and the Ph.D. degree from the University of Alberta in 2005. He is currently with the ECE Department at Utah State University, where he holds the rank of Associate Professor. Dr. Winstead's research interests include reliable wireless communication systems, implementation of error-correction algorithms, low-power electronics, fault-tolerant VLSI circuits, computation with noisy hardware and biochemical computation. In 2010, Dr. Winstead received the NSF Career award for research in low-energy wireless communication circuits. During 2013-2014, he was a Fulbright Visiting Professor at the Universite de Bretagne Sud (UBS) in Lorient, France. He is also a Senior Member of the IEEE and a member of the Tau Beta Pi engineering honor society.