Interest of low order moments on effective density estimation based on K-distribution
Résumé
An echographic signal often comes from a nonperiodic media with low density or clus-tered scatterers. Its envelope then deviates from pure Rayleigh statistics and the K- distribution is considered as a good model from which effective density estimators can be derived.") In previous work,(2) we studied the statistical limitations of a density estimator built on the second and fourth order moments of the envelope. In this work, we compare the perform- ances of this estimator with two others based on low order moments: order one and two in one case, and order and 1 in the other case. Results of numerical simulations are presented, where the rflines were computed for a large spectrum of densities : the envelope SNR ranges from 0.15 to 1.8. Bias and variance of the three density estimators studied were compared. The influence of an additive white noise in the bias term was also considered. On the experimental side, measurements were realized from four homogeneous foams having the same structure morphology but different pore size and volumic mass. For each foam, the effective density was estimated as well as other acoustical parameters, such as at- tenuation and backscattering coefficient. In both simulations and measurements, the estimator based on the moments of order one and two shows the best performances. This confirms the interest of low order moments den- sity estimators and shows that in good experimental conditions, the effective density can be considered as a pertinent and useful structure parameter.