Deep Modulation Recognition in an Unknown Environment
Lei Li, Qihang Peng, Pamela C. Cosman, L.B. Milstein · 2019
Deep modulation recognition has demonstrated high classification accuracy when a neural network is trained on large-scale datasets. However, when applied in an unknown environment where there are not any ground-truth labels in collected data, its performance can be significantly degraded. In this paper, we propose incorporating an adversarial discriminative neural network to adapt the deep modulation recognition to an unknown environment. Results show that, when the neural network is trained under an AWGN channel but applied under a frequency-selective Rayleigh fading channel, the adversarial network based domain adaptation can achieve comparable performance with that of the network trained with sufficiently large labeled data.