Specific Emitter Identification Based on Temporal Convolutional Network Sequence Processing
Chenyu Zhu, Liang Liu, Xiaoyan Peng
Abstract
Specific emitter identification (SEI) is a method to identify radiation sources using radio frequency fingerprints (RFF) during signal transmission. Since transient features are difficult to capture accurately, we often analyze steady-state fingerprints extracted from signals. In this letter, temporal convolutional networks (TCN) are utilized in the SEI field. It performs as well as recurrent neural network (RNN) on traditional sequential tasks, and it retains the excellent classification capabilities of convolutional neural network (CNN), so it can be used to process time series like signals. We can better extract and analyze fingerprints from the signal using the TCN modules instead of the original convolutional layers. Experimental results show that TCNs provide better recognition performance than methods that exclusively use traditional neural networks to process feature sequences.