Time

19.03.2018 kl. 13.00 - 19.03.2018 kl. 16.00

Description

Title
Network Coding Using Superregular Matrices for Robust Real-Time Streaming

Abstract
Streaming of audio and video has become a part of our everyday lives. Therefore, this work seeks to improve the quality of wireless audio streaming. Moreover, the main topic of this thesis is erasure correcting codes, that is, codes that are used to correct the erasures of entire packets in wireless networks. These codes may be used to make low delay wireless media streaming more robust, i.e., more tolerant towards erasures caused by errors somewhere along the network path.

In order for erasure correcting codes to be beneficial for low delay streaming, they should optimize the encoding delay, decoding delay, and loss probability. In this work, we study codes with a lower triangular structure, which minimizes the encoding delay, as opposed to dense codes, e.g., Reed-Solomon codes, where the source must collect all input symbols before the encoding can take place. The lower triangular structure has also proven favourable with respect to decoding delay since packets can be decoded on-the-fly. The symbol loss probability is then investigated for these lower triangular codes. The result of this investigation is two-fold. First, a framework for determining the symbol loss probability for any static linear erasure correcting code is composed. Second, in order for a systematic lower triangular code to have optimal decoding capabilities, the redundancy part of the encoding matrix must be superregular.

Using these erasure correcting codes on modern embedded platforms, such as loud-speakers, requires an efficient software implementation. To this end, this thesis documents a high-performance software library for encoding and decoding linear erasure correcting codes. This software library has proven very valuable for this research project, as it enabled large simulation campaigns and allowed for the implementation of a real-time multi-node wireless testbed.

Assessment Committee
Associate Professor Jimmy Jessen Nielsen, Aalborg University, Denmark (Chairman)
Professor Lars K. Rasmussen, KTH Royal Institute of Technology, Sweden
Professor Emina Soljanin, Rutgers School of Engineering, USA

Supervisors
Professor Jan Østergaard, Aalborg University, Denmark
DSP Specialist John Hammer Madsen, B&O, Denmark
Tech. Specialist Johnny Kudahl, B&O, Denmark

Moderator
Professor Søren Holdt Jensen, Aalborg University, Denmark

After the defence there will be a reception at Fredrik Bajers Vej 7, B4-211

Price

Free of charge

Host

Signal and Information Processing, Department of Electronic Systems

Address

Aalborg University, Fredrik Bajers Vej 7B3-104