AbstractAbstract
This work presents a study on the performance of instantaneous
frequency (IF) estimation of single-tone sinusoidal signals, and
IF estimation of FM signals under several types of colored noise
(CN). For single-tone sinusoidal signals, two main estimators
are considered: the maximum likelihood (ML) estimator using
the Fast Fourier Transform (FFT) with interpolated peak
estimation and the correlation method. Four statistical models
of colored noise are modelled and analyzed in this work
(Gaussian, Rayleigh, impulsive and a combination of impulsive
and Gaussian) with a study of their effect on frequency
estimation methods in term of estimation accuracy and
computational complexity presented. Also presented in this
work is a study of the performance of frequency estimation
methods under the same power and noise model. This work
lastly presents a study on the performance of instantaneous
frequency estimators for mono-component FM signals,
including linear and non-linear FM signals. Peak of a specific
Time-Frequency Distribution (TFD), the periodogram, has been
used for IF estimation of FM signals under colored noise.
Numerical results showed that the Quinn method is the best
estimator as compared with other FFT interpolated peak
estimation methods under colored noise in terms of minimum
mean squared estimation error, especially at high signal-to-noise
ratios. Simulation results showed also that Colored Rayleigh
noise is destructive for signals more so than colored Gaussian
noise and colored impulsive – plus- Gaussian noise.
Fourier approach can guess the frequency only in high SNR. Its
impact on IF estimation by correlation (corr) is worse.
Simulation results of instantaneous frequency estimators of FM
signals using TFD showed greater accuracy than with previous
methods.