Comparison of Analytic Signal Usage on Doppler Blood Flow Signals Estimation

Abstract

Spectral resolution may be enhanced by calculation of the analytic version of the signals processed. In this paper an open issue in the spectral estimation of Doppler ultrasound blood flow signals is addressed: when should the analytic signal representation be obtained and how to achieve such a representation. A comparison of the impact of two alternatives for achieving such a representation is reported. This is done by comparing the statistical performance of the short time Fourier transform on the estimation of mean frequency and bandwidth of simulated Doppler ultrasound blood flow signals. The comparison evaluates two approaches to obtain the analytic signal representation, one in the frequency domain -the common Fourier transform based approach - and one in the time domain - using a FIR filter with small latency that approximates the Hilbert transform. The results obtained show that the "analytic" signal should be used before the estimation process and that, in most cases, the use of the FIR filter diminishes the estimation error.