Parallel implementation of a Choi-Williams TFD for Doppler signal analysis

The spectral width and mean frequency of Doppler signals is used to measure and characterise lesion-induced flow disturbance. Conventional methods of spectral estimation based on the Fourier transform compromise the measurement accuracy since these methods implicitly assume signal stationarity during the signal window under analysis. New methods based on Time-Frequency Distribution (TFD), especially developed for non-stationary signal analysis, are now being introduced which avoid these limitations, at the cost of an increase in the computation burden required. This paper describes the parallel implementation of one TFD method, the Choi-Williams Distribution (CWD), in a transputer platform with 5 processors. The goal is to obtain a real time spectral analyser