2019/11/18 by Nuwanthika Rajapaksha, Rajapaksha, Nuwanthika, Nandana Rajatheva +3 · 1 citation
Computer Science · Engineering · #Antenna Design and Optimization #FOS: Computer and information sciences #FOS: Electrical engineering #Information Theory (cs.IT) #Machine Learning (cs.LG) #Signal Processing (eess.SP) #Speech and Audio Processing #Wireless Signal Modulation Classification #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1911.08009
openalex publication_date 2019/11/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
End-to-end learning of a communications system using the deep learning-based\nautoencoder concept has drawn interest in recent research due to its\nsimplicity, flexibility and its potential of adapting to complex channel models\nand practical system imperfections. In this paper, we have compared the bit\nerror rate (BER) performance of autoencoder based systems and conventional\nchannel coded systems with convolutional coding (CC), in order to understand\nthe potential of deep learning-based systems as alternatives to conventional\nsystems. From the simulations, autoencoder implementation was observed to have\na better BER in 0-5 dB Eb/N0 range than its equivalent half-rate\nconvolutional coded BPSK with hard decision decoding, and to have only less\nthan 1 dB gap at a BER of 10-5. Furthermore, we have also proposed a novel\nlow complexity autoencoder architecture to implement end-to-end learning of\ncoded systems in which we have shown better BER performance than the baseline\nimplementation. The newly proposed low complexity autoencoder was capable of\nachieving a better BER performance than half-rate 16-QAM with hard decision\ndecoding over the full 0-10 dB Eb/N0 range and a better BER performance\nthan the soft decision decoding in 0-4 dB Eb/N0 range.\n