Deep Ensemble Learning with Reconstruction Cancellation for Specific Emitter Identification

Abstract

Specific emitter identification (SEI) is an effective technique in the target identification field, which can recognize a specific transmitter by extracting the radio frequency (RF) fingerprints from the received signal. In this paper, we propose a novel SEI algorithm based on deep ensemble learning with reconstruction cancellation. Firstly, we demodulate the received signal to obtain the demodulation sequence. And then, we reconstruct an ideal waveform through the demodulation sequence. We design three kinds of characteristic representations both for the received signal and the corresponding ideal waveform. Meanwhile, we construct an ensemble learning model with a deep neural network to extract and fuse multiple features via data-driven strategies. Simulation results demonstrate that the proposed scheme achieves over 20% performance improvement learning than the single feature representation in a low signal-noise ratio.