Using sequence to sequence learning for digital BPSK and QPSK demodulation

Kalade, Sarunas and Crockett, Louise and Stewart, Robert (2018) Using sequence to sequence learning for digital BPSK and QPSK demodulation. In: IEEE 5G World Forum. IEEE, Piscataway, NJ. (In Press)

[img]
Preview
Text (Kalade-etal-5GWF2018-Using-sequence-to-sequence-learning-for-digital-BPSK)
Kalade_etal_5GWF2018_Using_sequence_to_sequence_learning_for_digital_BPSK.pdf
Accepted Author Manuscript

Download (430kB)| Preview

    Abstract

    In the last few years Machine Learning (ML) has seen explosive growth in a wide range of research fields and industries. With the advancements in Software Defined Radio (SDR), which allows more intelligent, adaptive radio systems to be built, the wireless communications field has a number of opportunities to apply ML techniques. In this paper, a novel approach to demodulation using a Sequence to Sequence (Seq2Seq) model is proposed. This type of model is shown to work effectively with PSK data and also has a number of useful properties that are not present in other machine learning algorithms. A basic Seq2Seq implementation for BPSK and QPSK demodulation is presented in this paper, and learned properties such as Automatic Modulation Classification (AMC), and ability to adapt to different length input sequences, are demonstrated. This is an exciting new avenue of research that provides considerable potential for application in next generation 5G networks.