Cerebral microemboli remain a high part of all vascular cerebral accidents in the western world. Thus detection, sizing and nature (particulate or gaseous) evaluation are three important tasks which are necessary to understand embolisms and to perform correct diagnostics and even therapy. The problem of detection has been widely reported in many works. Our aim was here to address the emboli characterization, sizing and nature evaluation. This characterization is based on EBR (embolus to blood ratio). EBR values were computed with two signal processing approaches: Fourier technique and parametric modeling. Emboli were simulated with acrylic particles ranging from 220 /spl mu/m to 500 /spl mu/m. With the smallest emboli, parametric approaches gave variance and standard deviation better than Fourier techniques. In a second time, we use the same multifrequency approach to find the frequencies triplet which assumed the best differentiation between gaseous and particulate emboli over a high range of diameters.