Bursty impulse noise detection by compressed sensing

Impulse noise (IN) is one of the major impairments for data transmission over power lines. For power line communications (PLC) systems with bandwidths in the high kHz to MHz range, IN occurs in bursts. As long as those bursts are sufficiently short compared to a signal-processing (e.g. coding) frame, there is hope to successfully mitigate IN. In this paper, we introduce a new IN mitigation technique that is based on the application of block-based compressed sensing (CS). It makes use of null-subcarriers in PLC orthogonal frequency division multiplexing (OFDM) transmission systems and the burst structure of IN to detect the location and the values of IN samples in the received signal. We also devise a semi-analytical error-rate performance evaluation for coded OFDM over IN channels, which allows insights into how CS-based IN detection can be used for improved reliability of transmission. Numerical results for typical PLC transmission settings demonstrate the efficacy of the proposed application of CS for IN detection.