TL;DR
This paper proposes a novel method for estimating the Nyquist Zone (NZ) index and chirp rate of Linear Frequency Modulated (LFM) signals intercepted by a Nyquist Folding Receiver (NYFR). By combining Random Sample Consensus (RANSAC) for initial estimation and Fractional Fourier Transform (FrFT) for precise chirp rate determination, the method achieves robust and accurate parameter estimation, even in low Signal-to-Noise Ratio (SNR) conditions.
What Problem Does This Paper Address?
Modern radar systems increasingly employ Low Probability of Intercept (LPI) signals to evade detection. These signals often have wide bandwidths and complex modulations. Capturing such signals with conventional receivers requires high-speed Analog-to-Digital Converters (ADCs), which are expensive and power-hungry. The Nyquist Folding Receiver (NYFR) offers a solution by undersampling the wideband radio frequency (RF) signal before digitization, effectively folding it into a lower frequency band. However, this folding process introduces a hybrid modulation with the receiver's Local Oscillator (LO), making parameter estimation (like chirp rate and the specific "Nyquist Zone" the signal landed in) challenging. Existing methods struggle with accuracy, especially at low SNRs, or are not well-suited for the complex output of the NYFR. This paper aims to provide a robust and accurate method for estimating key LFM signal parameters from the NYFR output.
Methods
The proposed method involves a two-stage approach:
- Nyquist Zone (NZ) and Rough Chirp Rate Estimation:
- The authors first analyze the time-frequency characteristics of the LFM signal after it has been processed by the NYFR. This output is a hybrid of the original LFM signal and the NYFR's sinusoidal frequency modulation (SFM) LO.
- Least Squares (LS): Initially, LS is used to fit lines to the upper and lower bounds of the signal's time-frequency spectrogram. The slope of these lines provides an initial estimate of the chirp rate, and the interval between them relates to the NZ.
- Random Sample Consensus (RANSAC): Recognizing that LS is sensitive to outliers (especially at low SNRs), the paper employs RANSAC. RANSAC iteratively selects random subsets of "characteristic points" from the time-frequency plane to fit a linear model. It then identifies "inliers" that fit the model well and discards "outliers." This robust approach significantly improves the accuracy of estimating the upper and lower bounds, and consequently, the NZ and a rough chirp rate, even when the SNR is as low as -5 dB.
- Precise Chirp Rate Estimation using Fractional Fourier Transform (FrFT):
- Once the NZ is accurately determined, the SFM component of the LO is effectively removed from the signal. This leaves a cleaner LFM signal.
- Fractional Fourier Transform (FrFT): The FrFT is a generalization of the Fourier Transform that can transform a signal into an intermediate domain between time and frequency. For an LFM signal, the FrFT exhibits a peak at a specific "order" (or rotation angle) that is directly related to the signal's chirp rate.
- Fast Search Method for Optimal FrFT Order: To efficiently find this optimal order, the paper proposes a fast search algorithm. Instead of a brute-force planar search, it uses a stepwise refinement approach, significantly reducing computational complexity while maintaining accuracy. This allows for precise estimation of the chirp rate.
Datasets
Not discussed in the paper. The evaluation relies on simulated LFM signals with varying SNRs.
Key Insights and Ideas
- RANSAC for Robustness: The paper highlights the critical role of RANSAC in overcoming the limitations of traditional LS fitting for time-frequency analysis, particularly in noisy environments. Its ability to identify and utilize "inliers" makes it ideal for extracting signal features from the complex NYFR output.
- FrFT's Chirp Rate Sensitivity: The core idea that the FrFT's peak location is directly tied to the LFM signal's chirp rate is leveraged effectively. This provides a powerful tool for precise chirp rate estimation once the signal is isolated.
- Efficient FrFT Order Search: The proposed fast search method for the optimal FrFT order is a significant contribution, addressing the computational burden of traditional FrFT-based parameter estimation. This makes the overall method more practical for real-time applications.
- Synergy of RANSAC and FrFT: The paper demonstrates a powerful synergy between RANSAC and FrFT. RANSAC provides the robust initial estimation of the signal's location (NZ) and a rough chirp rate, enabling the subsequent precise chirp rate estimation by FrFT.
Key Results
- The proposed RANSAC and FrFT method achieves high accuracy in estimating both the Nyquist Zone (NZ) index and the chirp rate of LFM signals intercepted by an NYFR.
- NZ Determination: The probability of correct NZ determination exceeds 90% for SNRs greater than or equal to -9 dB and reaches 100% for SNRs above -7 dB. This is superior to other considered methods like peak search and autocorrelation.
- Chirp Rate Estimation: The Normalized Root Mean Square Error (NRMSE) for chirp rate estimation is significantly lower for the RANSAC and FrFT method, especially at lower SNRs (e.g., below -5 dB). The method is stable and performs well even at -8 dB SNR.
- The fast search method for the FrFT optimal order significantly reduces processing time compared to traditional methods, with accuracy improving exponentially with approximation orders while processing time grows linearly.
Limitations and Future Work
- Single LFM Signal Assumption: The paper explicitly states that for ease of deduction, it was assumed that the received signal contained only a single LFM signal and that its frequency range did not cross NZ junctions.
- Future Work: The authors suggest that the processing of multiple signals within the NYFR, and scenarios where the LFM signal spans multiple NZs, should be investigated further. This would involve extending the current method to handle more complex signal environments.
Summary updated