Joint Coding and Modulation in the Ultra-Short Blocklength Regime for Bernoulli-Gaussian Impulsive Noise Channels Using Autoencoders

This paper develops a joint coding and modulation scheme for end-to-end communication system design using an autoencoder architecture in the ultra-short blocklength regime. Unlike the classical approach of separately designing error correction codes and modulation schemes for a given channel, the approach here is to learn an optimal mapping directly from messages to channel inputs while simultaneously learning an optimal mapping directly from channel outputs to estimated messages. The block error rate (BLER) of this approach is compared against classical short blocklength linear block codes with binary phase shift keying (BPSK) modulation in additive white Gaussian noise (AWGN) and Bernoulli-Gaussian impulsive noise (BGIN) channels. For AWGN channels, numerical results show that the autoencoder can achieve better BLER performance than BPSK modulated Hamming codes with maximum likelihood decoding. For BGIN channels, numerical results show the autoencoder achieves uniformly better BLER performance than conventional block codes with BPSK modulation, even with impulsive noise mitigation techniques such as blanking and clipping. The proposed architecture is general and can be modified for comparison against other block coding schemes and higher-order modulations.