Convolutional Autoencoder for Compressive Symbol Detection

摘要

This paper deals with a problem of digital communication which samples a signal compressively below its symbol rate and restores the original message symbols. We propose a method to optimize the sensing and restoration modules by designing them as the encoder and decoder of a convolutional autoencoder. The network learns end-to-end through simulation data. Our architecture has an efficient mechanism to solve this compressive symbol detection problem and it shows strong performance and fast speed empirically.