Chaos-Based Space-Time Trellis Codes With Deep Learning Decoding

Abstract

In this brief we propose a space-time trellis code scheme based on three-dimensional chaotic attractors. The chaotic trajectories are represented by the symbolic dynamics generated by a labeled Poincaré section and are transmitted by multiple antennas, defining a chaos-based space-time trellis code (CB-STTC). This code is defined by a finite state encoder that maps information sequences to restricted sequences satisfying the dynamics of the attactor. We also propose a neural network architecture capable of learning how to decode the CB-STTC. Finally, the frame error rate of the proposed CB-STTC is analyzed with maximum likelihood and neural network decoding.