基于时频表征优化与网络重构的宽带信号检测识别方法
By dividing the time-frequency characteristics of broadband signals into an energy-phase joint domain and an energy domain, and reconstructing them using ResNet50 and CV-ResNet50 backbone networks, as well as Sparse R-CNN and DINO models, the problem of discarding phase information in existing methods is solved, and more efficient broadband signal detection and recognition are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AIR FORCE UNIV PLA
- Filing Date
- 2026-05-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing broadband signal detection and recognition methods discard phase information when processing complex time-frequency spectra, leading to a decrease in the performance of modulation recognition models, and the impact of different signal representation methods on task performance has not been thoroughly explored.
By dividing the ten time-frequency feature representations into the energy-phase joint domain and the energy domain, and inputting them into the ResNet50 and CV-ResNet50 backbone networks, and combining the Sparse R-CNN model and the DINO model for target detection, a common basis loss is defined for model update, thereby achieving accurate recognition of the detected signal.
It effectively adapts to different styles of time-frequency feature representation, improves the accuracy and robustness of signal detection, overcomes the information loss caused by traditional power conversion, and improves the performance of broadband signal detection and recognition.
Smart Images
Figure CN122247540B_ABST