LDPC Codeword Size Determination Using Convolutional Neural Networks

This paper discusses the design and performance of a forward error correction (FEC) code classification system that is used to determine the size of an unknown codeword from a stream of bits. The classification system is a deep neural network that is trained and tested on half rate low density parity check (LDPC) codes. Tests were performed on streams of codewords using codes of up to 250 different sizes. The CNN based classifier performs very well.