Zero-shot Learning with Cross-Layer Neural Network for Emitter Pattern Recognition
- Resource Type
- Conference
- Authors
- Zhang, Zilin; Li, Yan; Bai, Jinliang
- Source
- 2019 IEEE International Conference on Signal, Information and Data Processing (ICSIDP) Signal, Information and Data Processing (ICSIDP), 2019 IEEE International Conference on. :1-5 Dec, 2019
- Subject
- Aerospace
Communication, Networking and Broadcast Technologies
Signal Processing and Analysis
Emitter Pattern Recognition
Zero-shot Learning
Cross-Layer Neural Network
- Language
The existing emitter pattern recognition methods depend on a large number of labeled samples and are unable to handle unknown samples. Zero-shot Learning (ZSL) can migrate from source classes to target categories by learning a common embedding space, thus realizing the generalization to unknown samples. In this paper, a novel Cross-Layer Neural Network (CLNN) is proposed that integrates different embedding methods into an end-to-end deep learning architecture. The experimental results demonstrate that the proposed method can achieve excellent performances in the presence of unseen radar patterns with either new feature combinations or new feature ranges.