Comparison of parametric and nonparametric spectral estimation of continuous Doppler ultrasound shift waveforms

As an alternative to the fast Fourier transform (FFT) spectral analysis approach, an autoregressive (AR), a moving average (MA) and an autoregressive moving average (ARMA) model based spectral estimators were used to process the continuous wave Doppler ultrasonic aortic flow signal. The spectrogram of each method was subsequently plotted. This study was undertaken to compare the maximum frequency shift envelope of each of the spectrograms and to find the best parametric model suitable for this investigation. The Doppler signal was subdivided into overlapping windowed data sets. The FFT estimate was performed after applying a Hamming window to each data record. The AR model was constructed and solved using Yule-Walker (Y-W) equations. The MA model was estimated from a truncated higher order AR model and solved likewise using Y-W equations. The ARMA estimate was obtained by evaluating the AR parameters first, using the extended Y-W equations. The AR parameters were then used to obtain a filtered MA sequence of the data and subsequently estimate the MA part as described previously. Overall, the use of parametric model based approach provided cleaner spectrographs