Frequency estimation is transformed to a pattern recognition problem, and a least squares support vector machine (LS-SVM) estimator is derived. The estimator can work efficiently without the need of statistics knowledge of the observations, and the estimation performance is insensitive to the carrier phase. Simulation results are presented showing that proposed estimators offer better performance than traditional Maximum Likelihood (ML) estimator at low SNR, since classification-based method does not have the threshold effect of nonlinear estimation.
Paper
Full text
Using Least Squares Support Vector Machines for Frequency Estimation
Semantic Scholar · Computer Science · 2010
Abstract
Frequency estimation is transformed to a pattern recognition problem, and a least squares support vector machine (LS-SVM) estimator is derived. The estimator can work efficiently without the need of statistics knowledge of the observations, and the estimation performance is insensitive to the carrier phase. Simulation results are presented showing that proposed estimators offer better performance than traditional Maximum Likelihood (ML) estimator at low SNR, since classification-based method does not have the threshold effect of nonlinear estimation.