Skip to main navigation Skip to search Skip to main content

A noise-robust FSK modulation classification method via local mean decomposition and optimized residual network

  • Yujie Xiao
  • , Li Li
  • , Weichao Yang
  • , Dawei Wang
  • , Ke Ren
  • , Yi Jin
  • Northwestern Polytechnical University Xian

Research output: Contribution to journal › Conference article › peer-review

Abstract

A Local Mean Decomposition (LMD)-enhanced ResNet framework is proposed for FSK classification. Local mean decomposition discards the highest-frequency noise-dominated product function; the remaining nine PFs are stacked into a 2-D feature map. An auxiliary branch in ResNet50 element-wise adds shallow texture (Stage 2) to deep semantics (Stage 5), enabling multi-scale fusion. Experiments demonstrate ≥90% accuracy for 2/4/8-FSK at SNR ≥ 0 dB, outperforming existing low-SNR methods.

Original languageEnglish
JournalJournal of Physics: Conference Series
Volume3235
Issue number1
DOIs
StatePublished - 2026
Event2026 5th International Conference on Electronic Information and Communication Engineering (EICE 2026) - Sanya, China
Duration: 30 Jan 2026 → 1 Feb 2026

Fingerprint

Dive into the research topics of 'A noise-robust FSK modulation classification method via local mean decomposition and optimized residual network'. Together they form a unique fingerprint.

Cite this