A symbol flipping decoder for NB-LDPC relying on multiple votes

In this paper, we present an algorithm to decode non-binary LDPC (NB-LDPC) codes, inspired from very-high throughput symbol-flipping decoders that have been proposed recently. Usually, the symbol-flipping decoders suffer from a non-negligible performance degradation compared to soft-decision NB-LDPC decoders. Our improved decoder makes use of a list of syndrome computations instead of a single one, and builds soft information at the symbol node input by assigning votes with different weights to the elements of the list. We show by simulations that using multiple votes results in better performance, while still maintaining the high throughput feature.