An Improved Modulation Recognition Algorithm Based on Fine-Tuning and Feature Re-Extraction

• DOI: 10.3390/electronics12092134
• Publication Date: 2023-05-08T06:03:31Z
• Authors: Yibing Wang; Liang Zhou; Zhutian Yang; Longwen Wu; Zhendong Yin; Yaqin Zhao; Zhilu Wu
• Keywords: Modulation (music); Feature (linguistics); Fine-tuning; Basis (linear algebra)

Modulation recognition is an important technology in wireless communication systems. In recent years, deep learning-based modulation algorithms, which can autonomously learn features and achieve superior performance compared with traditional have emerged. Yet, there are still certain limitations. this paper, aiming at addressing the issue of poor low signal-to-noise ratios (SNRs) inability to effectively distinguish among all types, we propose optimization scheme for based on fine-tuning feature re-extraction. proposed scheme, network firstly trained signals high SNRs; then, fine-tuned untrained SNRs. Finally, basis learned by network, deeper enhanced discriminability confused types obtained using The simulation results demonstrate that maximize neural easily Notably, average accuracy was 91.28% within SNR range −8 dB 18 dB, improvement 8% 17% comparison four existing schemes.