| MS Seminar


Name of the Speaker: Mr. ARPIT KUMAR (EE22S086)
Guide: Dr. Arun Pachai Kannu
Venue: ESB-244 (Seminar Hall)
Online meeting link: http://meet.google.com/nvp-xzkb-dic
Date/Time: 10th June 2025 (Tuesday), 4:00 PM
Title: Sparse Regression Codes for Coherent multi-antenna Channels for multi- user communication.

Abstract :

e study sparse regression codes (SPARC) for short and moderate block lengths over flat fading channels for multi-user communication with multiple receive antennas. We consider a coherent communication scenario where the receiver has perfect channel state information.For the single-user case, we illustrate that greedy algorithms such as orthogonal matching pursuit (OMP) have limitations on the code rate supported by SPARC when the block length increases. To overcome this limitation, we introduce two new decoders, OMP with replacement (OMP-R) and maximum likelihood matching pursuit (MLMP) with parallel search. We study the block error rate performance of SPARC over flat fading channels in the moderate block length regime and show that the proposed decoders outperform the existing sparse signal recovery algorithms, OMP and CoSAMP. In addition, MLMP, with much lower complexity, performs very close to the approximate message passing algorithm, which has been shown to achieve the channel capacity for large block lengths.We then extend our study to the multi-user setting, where multiple users transmit simultaneously over a shared multi-antenna channel. This introduces new challenges such as inter-user interference and joint decoding under sparsity constraints. To tackle this, we developed a multiuser MLMP decoder and also introduced a parallel search-based variant of it, referred to as Parallel MLMP (P-MLMP), to enhance decoding performance and scalability. We compared it against existing channel coding techniques such as orthogonal multiple access polar codes. Our simulations demonstrate that P-MLMP offers robust decoding performance in multiuser environments, achieving near-optimal performance while maintaining manageable complexity. These results highlight the scalability of SPARC-based schemes to multiuser systems and reinforce their suitability for modern wireless applications.