Deep Ensemble Learning for Automatic Modulation Classification

Abstract

Automatic modulation classification (AMC) plays an increasingly vital role in cognitive radio (CR), cognitive electronic warfare, and other areas. It aims at classifying the modulated modes of the received signals accurately and provides a guarantee for the subsequent detailed parameter identification. Deep learning (DL) methods allow the computer to automatically learn the pattern features and integrate features into the process of building the model, thereby reducing the incompleteness caused by artificial design features. At the same time, the DL methods have been applied in the AMC field as its powerful ability to process complex data and have achieved excellent performance in recent years. In this paper, we propose a deep ensemble learning AMC network, which uses a multi-model ensemble method to fuse multiple DL features. Specifically, different DL models are integrated by ensemble learning, which enhances the learning ability of the single model. With the proposed ensemble model trained on a measured wireless signal dataset, we conclude that the ensemble structure of Inception and CLDNN can fuse spatial features and temporal features, and achieve state-of-the-art performance in AMC tasks. Besides, the impact of the inphase/quadrature (I/Q) sample-length on wireless signals is further investigated, and find that the classification accuracy of the deep ensemble model is improved by 0.7% to 10% compared to the single model under various sample-length. Simultaneously, we visualize convergence clustering with t-distributed stochastic neighbor embedding (t-SNE), and the visualization results prove that the deep ensemble model has a stronger clustering ability than a single model.

Generative UAP attacks against deep‐learning based modulation classification

“…For example, in cognitive radios where the licensed primary users and the unlicensed secondary users coexist by sharing some spectrum resources, spectrum sensing and modulation classification are the requisite for the secondary users to communication successfully using unlicensed spectrum in an opportunistic manner. To achieve a better performance, investigations have been conducted to apply DL‐based algorithms to improve the accuracy of modulation classification [15, 16]. It was found, however, that the DL‐based algorithms for modulation classifications are also vulnerable to adversarial perturbations [17, 18].…”

“…validated that the recognition of a simple CNN network trained on RML2016.10a can achieve an accuracy of more than 90% for most modulated signals. In [15], Ramjee et al. conducted a work using different DNN network structures, for example, RNN, to achieve a higher classification accuracy than CNN.…”