LFM Signal Sources Classification Based on Self-supervised Learning
Tianqi Yang, Siya Mi · Progress In Electromagnetics Research Letters · 2023
Linear Frequency Modulation (LFM) signals are widely used in radar and sonar technology.Many applications are interested in determining the source of an LFM signal.In recent years, the rapid development of machine learning has facilitated research in various fields, including signal recognition.The neural networks can extract the implicit features of the signals, which can help the system to sort and recognize the signal sources quickly and accurately.High performance of neural networks requires large amounts of high-quality labeled data.However, it is difficult and expensive to obtain a large amount of high-quality labeled data.Simultaneously, some features will be lost during data preprocessing, and feature extraction and classification tasks will be inefficient.The self-supervised network is proposed in this paper for pre-training the signal waveform and fine-tuning the classification with a small amount of labeled data.The proposed method can extract more signal waveform features, save labeling costs, and has higher precision.This method can provide up to 99.7% recognition accuracy at 20 dB.