Performance of error correction based on Compressed Sensing

Abstract

In this paper, the real-valued Compressed Sensing (CS)-based error correction scheme which was proposed by Candes and Tao is evaluated against the background of channel coding theory. Substantial differences to conventional linear block codes over Galois fields are reviewed and interpreted. A unified channel model is introduced. It enables performance evaluation of more realistic almost sparse channels, which are of interest for communication applications. It is shown by simulations that certain variations of the CS based error correction scheme are suitable for special channels. Additionally, explanations for the noise sensitivity of CS based error correction schemes are given.