Abstract
Frequency estimation is a fundamental problem in many areas. The well-known A&M and its variant estimators have established an estimation framework by iteratively interpolating the discrete Fourier transform (DFT) coefficients. In general, those estimators require two DFT interpolations per iteration, have uneven initial estimation performance against frequencies, and are incompetent for small sample numbers due to low-order approximations involved. Exploiting the iterative estimation framework of A&M, we unprecedentedly introduce the Padé approximation to frequency estimation, unveil some features about the updating function used for refining the estimation in each iteration, and develop a simple closed-form solution to solving the residual estimation error. Extensive simulation results are provided, validating the superiority of the new estimator over the state-the-art estimators in wide ranges of key parameters.
Introduction
Frequency estimation of single-tone signals is a cornerstone of signal processing, essential for radar, vehicular sensing, and wireless communications. A popular class of estimators, such as the A&M estimator, works by iteratively interpolating Discrete Fourier Transform (DFT) coefficients to refine frequency estimates. However, these methods face significant hurdles: high computational costs per iteration, poor performance when the number of samples ($N$) is small, and uneven estimation bias across the frequency spectrum.
This paper introduces a novel frequency estimator that replaces traditional first-order Taylor Series approximations with Padé Approximations (PA) within an iterative DFT-interpolation framework. The result is a high-precision estimator that requires fewer computations and handles small sample sizes significantly better than state-of-the-art methods.
Methodology
The authors build upon the iterative framework of the A&M estimator but introduce a sophisticated mathematical upgrade to the "updating function" used to calculate residual error.
- Padé Approximation (PA): Instead of a linear Taylor Series, the method uses a Padé Approximation of degree (3,3). A PA represents a function as a ratio of two polynomials, which is far more effective at capturing the non-linear behavior of the sinc-like functions found in DFT ratios.
- Symmetry Exploitation: The updating function is proven to be monotonic with odd symmetry. These properties are used to zero out even-power coefficients in the PA, simplifying the derivation without sacrificing accuracy.
- Closed-Form Solution: By setting the PA equal to the measured DFT ratio, a cubic equation for the frequency error is derived. This can be solved exactly using the standard cubic formula, providing a direct, non-linear update.
- Efficiency: A "sign test" determines the initial frequency region, allowing the estimator to start very close to the true value with only a single interpolation in the first step. Total complexity is reduced as the algorithm typically converges in just two iterations.
Performance Analysis
The proposed method was validated through extensive Monte-Carlo simulations, comparing it against the original A&M, Generalized A&M (GAM), and Hybrid A&M and q-Shift Estimator (HAQSE).
Superiority in Small Sample Regimes
One of the primary weaknesses of traditional estimators is their reliance on approximations that only hold for large $N$. The PA-based approach maintains high accuracy even when $N$ is very small, a common scenario in spatial frequency estimation for antenna arrays.
As shown in Figure 3, the proposed estimator (marked with triangles) stays significantly closer to the Cramér-Rao Lower Bound (CRLB) across varying sample sizes. For $N=8$ (where $\log_2 N = 3$), the method achieves an MSE roughly 6.74% lower than A&M and GAM after the first iteration.
Uniformity Across the Frequency Spectrum
Traditional estimators often suffer from "ripples" in accuracy depending on where the true frequency lies relative to DFT bins. The higher-order accuracy of the Padé Approximation eliminates these performance dips.
Figure 5 demonstrates that the proposed method provides a remarkably uniform MSE across the entire range of fractional frequency offsets $\delta$. This ensures reliable performance regardless of the signal's specific frequency.
Conclusion
By integrating Padé Approximations into the iterative DFT interpolation framework, this research provides a frequency estimator that is both more accurate and more computationally efficient than previous state-of-the-art methods. Its ability to reach the CRLB in just two iterations and its robustness at small sample sizes make it highly suitable for real-time, high-precision applications in modern sensing and communication systems.
Summary updated