Options
Information optimum design of discrete LDPC decoders for irregular codes
Citation Link: https://doi.org/10.15480/882.1710
Publikationstyp
Master Thesis
Publikationsdatum
2018
Sprache
English
Author
Advisor
Referee
Title Granting Institution
Technische Universität Hamburg
Place of Title Granting Institution
Hamburg
Examination Date
2017-03
Institut
Citation
Masterarbeit am Institut für Nachrichtentechnik der TUHH (2018)
Powerful forward error correction is inevitable for every modern communication system. However, leveraging all error correction capabilities offered by channel codes like low density parity check (LDPC) codes, introduces the burden of computationally demanding decoding at the receiver. Especially in power or latency restricted scenarios, efficient decoding algorithms are of crucial importance. In this thesis a novel decoding approach combining ideas from information theory and machine learning was developed, resulting in a so-called information-optimum LDPC decoder. In particular, the information bottleneck method was used to obtain very compact representations of the exchanged beliefs while performing belief propagation decoding. To generalize the decoding approach, to be applicable also for practically relevant irregular LDPC codes another technique called message alignment was devised. As a result, LDPC decoders can be constructed which pair very low-complexity and near-optimum performance (0.2dB performance degradation). Computer simulations for irregular LDPC codes from the IEEE 802.11 (WLAN) and DVB-S2 standard proof the applicability of the developed decoders also in standardized communication systems.
Schlagworte
LDPC codes
information theorie
information bottleneck method
machine learning
DDC Class
620: Ingenieurwissenschaften
Loading...
Name
Masterthesis_Maximilian_Stark.pdf
Size
1.07 MB
Format
Adobe PDF