Parallel Computing in Time-Frequency Distributions for Doppler Ultrasound Blood Flow Instrumentation

Resumen

Doppler blood flow spectral estimation is a technique for non-invasive cardiovascular disease detection. Blood flow velocity and disturbance may be determined by measuring the spectral mean frequency and bandwidth respectively. Typical methods for spectral estimation utilize Fourier Transform-based algorithms to estimate the spectral response of a signal. This current practice suffers from poor frequency resolution when estimating non-stationary signals. The work presented here describes some alternative methods based on time-frequency distributions from a Cohen´s class point of view. Four distribution cases are evaluated: Wigner Ville, Choi Williams, Bessel and Born Jordan. A discrete distribution is formulated for each case and a criterion for determining the interaction between the spectral components of the signal is also given. Simplified discretized expressions for the implementation of distributions are formulated, these leading to a reduction of the computations realized when comparing to original definitions. A general parallel approach for the computation of the distributions is proposed, implemented and assessed using a parallel DSP-based system. Results are applied to the development of a real-time spectrum analyzer for Doppler blood flow instrumentation.