Estimation of the blood Doppler frequency shift by a matching pursuit algorithm

The diagnosis of arterial occlusive disease often depends on the Doppler spectrum analysis. We can normally use the short-time Fourier transform (STFT) to compute the time-frequency representation (TFR) of the Doppler blood flow signal. This method uses a fixed time-frequency window, making it inaccurate to analyze signals with relatively wide bandwidths that change rapidly with time. In order to estimate the Doppler frequency shift more accurately, even when the temporal flow velocity is rapid (high non-stationarity), we propose to use a modified version matching pursuit (MP) with stochastic dictionaries to estimate the time frequency representation of Doppler blood flow signals for extracting the mean frequency shift. Results show that the modified MP method can provide more accurate mean frequency waveforms than the STFT does.