Estimating the frequency of a noisy sinusoid by linear regression
IEEE Transactions on Information Theory
Semi-Supervised Clustering of Corner-Oriented Attributed Graphs
HIS '06 Proceedings of the Sixth International Conference on Hybrid Intelligent Systems
Computationally efficient estimation of sinusoidal frequency at low SNR
ICASSP '96 Proceedings of the Acoustics, Speech, and Signal Processing, 1996. on Conference Proceedings., 1996 IEEE International Conference - Volume 05
| Hi-index | 0.00 |
Aclassical problem in signal processing is accurate estimationof fundamental frequency/periodicity of periodic signals at lowSNR. Typically, researchers address the estimation problem, assumingthat the signal environment is a sum of sinusoids in white Gaussiannoise. If the signals and noise are pulsed, the situation ismuch more complex since normal FFT based methods result in spectrawhich are sums of harmonic structures. Sorting radar signalscan be especially difficult since there may be many pulsed signalspresent in a low SNR impulsive noise environment. In this paper,a method equivalent to integration along a hyperbola on the Wignerdistribution is presented. This transform, which is closely relatedto both the Fourier transform and the correlation function, hasthe property that a periodic signal produces an expected non-zerocomplex-valued bulge at only the fundamental. The phase, magnitudeand position of the correlation bulge are sufficient to characterizethe time-domain pulse train. Finally, a simple super-resolutionmethod is presented which may be used to refine the fundamentalfrequency/period estimate.