Digital methods for spectral analysis of noisy nonstationary distorted signal

The paper deals with the estimation of characteristic parameters of digitised non-stationary noisy signal. The measurement accuracy of instantaneous frequency and amplitude of the time-varying signals are carefully investigated. First, a suitable discrete Wavelet transform algorithm and a modified discrete interpolated short-time Fourier transform are analysed and compared for analysis of heavily non-stationary signals, with particular reference to their noise rejection and systematic errors caused by variations of the signal’s instantaneous parameters. Then, an accurate estimation of the instantaneous frequency and amplitude is obtained even if multicomponent signals are considered. Finally, the simplified expressions of systematic and of casual errors in terms of the used filter, the parameters of the signal and the characteristics of noise are carried out.