Deterministic Compressed Sensing with Power Decoding for Complex Reed-Solomon Codes

Complex Reed-Solomon (CRS) codes can be used for deterministic Compressed Sensing (CS). In order to improve the recovery capabilities, power decoding for CRS codes is investigated. We discuss the robustness of known error locator algorithms against the numerical instabilities arising from floating point calculations. A novel noise robust algorithm inspired by Generalized Minimum Distance decoding is introduced. Its performance is evaluated and compared to Orthogonal Matching Pursuit which is a well known CS algorithm.