|Publisher DOI:||10.1109/ISTC.2018.8625342||Title:||A Discrete Information Bottleneck Receiver with Iterative Decision Feedback Channel Estimation||Language:||English||Authors:||Lewandowsky, Jan
|Issue Date:||23-Jan-2019||Source:||International Symposium on Turbo Codes and Iterative Information Processing, ISTC (2018-December): 1-5 (2019-01-23)||Journal or Series Name:||International Symposium on Turbo Codes and Iterative Information Processing, ISTC||Abstract (english):||The application of the Information Bottleneck method allows for an innovative approach to quantized, but information-optimum signal processing. This method originates from machine learning, but also offers an unconventional and efficient design approach for communication receivers. In the resulting receivers all signal processing degenerates to simple lookup operations in tables. These lookup tables are designed to maximize the relevant information. This paper presents a novel iterative Information Bottleneck receiver for block fading channels which uses iterative decision feedback for information-optimum channel estimation. Detection, channel decoding and channel estimation only use lookup tables constructed with the Information Bottleneck method. We compare the proposed receiver to conventional iterative receivers. Our findings show that the proposed Information Bottleneck receiver offers performance extremely close to that of a conventional double precision receiver. Moreover, it outperforms a fixed point receiver, even with entirely mismatched lookup tables. © 2018 IEEE.||Conference:||10th IEEE International Symposium on Turbo Codes and Iterative Information Processing||URI:||http://hdl.handle.net/11420/2234||ISBN:||978-153867048-4||ISSN:||2165-4700||Institute:||Nachrichtentechnik E-8||Type:||InProceedings (Aufsatz / Paper einer Konferenz etc.)|
|Appears in Collections:||Publications without fulltext|
Show full item record
checked on Sep 29, 2020
Add Files to Item
Note about this record
Items in TORE are protected by copyright, with all rights reserved, unless otherwise indicated.