EE Seminar: Polar Coded Modulation
~~
Speaker: Peer Berger,
M.Sc. student under the supervision of Prof. Simon Litsyn
Wednesday, October 21, 2015 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering
Polar Coded Modulation
Abstract
Polar codes are the first provably capacity achieving codes for the class of binary-input symmetric discrete memoryless channels with low encoding and decoding complexity.
It is clear that polar codes are very attractive, but up until now most researches considered only the BPSK scenario. Unlike in the BPSK case - where code construction is the major concern, one can divide the general polar coded modulation problem into three major parts - code construction, constellation choosing and decoder choosing. Moreover, in the polar coded modulation case, performance refers to much more than BLER/BER.
This seminar will deal with aspects related to polar coded modulation. It will provide a set of tools and observations that improves the design of currently used polar coded modulation schemes (code construction for a certain constellation, kernel and decoder) that meets ones requirements (error probability, complexity and PAPR). We will do so by answering all three questions above, presenting new ideas, developing new criterions, using new modifications, and most important achieving better performance than known results.
In this seminar, two new labeling choosing criterions will be presented for arbitrary constellation, kernel and decoder, named Genie-Aided Gaussian-Approximation Union-Bound Criterion and Genie-Aided Single-Constellation-Polar-Code Criterion. A significant BLER performance improvement will be shown using those criterions. An error probability and PAPR reduction will be shown using different constellation sizes, APSK schemes and multi-dimensional constellations schemes. A PAPR reduction using APSK schemes, and an error probability and PAPR reduction using multi-dimensional constellations schemes will be presented. A new SCL-SC MLPC decoder that gives similar results to the SCL MLPC decoder with less complexity and a new Mixed-Kernels-MLPC scheme are being used in order to attain a better performance-complexity tradeoff. Finally, a new BIMLPCM decoder concept that potentially has better results than the capacity achieving MLPC decoder will be presented.
