Scholarlink

<p>This study benchmarks feature extraction techniques and classifier architectures for the Specific Emitter Identification (SEI) of real-world FM signals, specifically addressing the trade-off between accuracy and computational efficiency. By evaluating Support Vector Machines (SVM) against a 1D-CNN baseline across time, frequency, and hybrid domains, we demonstrate that time-domain features significantly outperform frequency-domain counterparts (micro-averaged AUC: 0.9009 vs. 0.8428). Furthermore, a hybrid fusion framework delivers peak performance (AUC: 0.9106). Crucially, the hybrid SVM-RBF model achieves accuracy comparable to the deep learning baseline with substantially lower computational demands, validating its suitability for resource-constrained signal intelligence applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid time-frequency feature fusion for FM transmitter identification: physical interpretability and performance benchmarking

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Narathep Phruksahiran

摘要

This study benchmarks feature extraction techniques and classifier architectures for the Specific Emitter Identification (SEI) of real-world FM signals, specifically addressing the trade-off between accuracy and computational efficiency. By evaluating Support Vector Machines (SVM) against a 1D-CNN baseline across time, frequency, and hybrid domains, we demonstrate that time-domain features significantly outperform frequency-domain counterparts (micro-averaged AUC: 0.9009 vs. 0.8428). Furthermore, a hybrid fusion framework delivers peak performance (AUC: 0.9106). Crucially, the hybrid SVM-RBF model achieves accuracy comparable to the deep learning baseline with substantially lower computational demands, validating its suitability for resource-constrained signal intelligence applications.