A radar signal intelligent recognition and parameter estimation method based on time-frequency mask
By constructing the TFM-RSPEN network, the problems of accuracy and versatility in radar signal identification and parameter estimation under noisy scenarios were solved, achieving high-precision identification and parameter estimation of radar signals with various modulation types, and improving the effect of radar signal processing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies for radar signal identification and parameter estimation in noisy scenarios suffer from poor universality and low accuracy in parameter estimation, especially for non-stationary signals and radar signals with multiple modulation types, making it difficult to achieve high-precision identification and estimation.
A radar signal intelligent identification and parameter estimation method based on time-frequency masking is adopted. By constructing a TFM-RSPEN network, signal features are extracted using adaptive thresholding and multi-scale convolutional submodules. The signal modulation type is identified and parameters are estimated by combining a time-frequency-parameter space mapping module. Weighted cross-entropy loss and Dice function are used to optimize network training.
It achieves accurate identification and high-precision parameter estimation of various signal modulation types in noisy scenarios. In particular, the identification accuracy is close to 100% under low signal-to-noise ratio conditions, and the bandwidth and pulse width estimation errors are less than 0.1, which is significantly better than traditional methods.
Smart Images

Figure CN122430795A_ABST