A reactor noise approach has been successfully performed at the IPEN/MB-01 research reactor facility to determine experimentally the effective delayed neutron parameters i and i in a six-group model. The method can be considered a novel one because it exploits the very low-frequency domain of the spectral densities. The proposed method has some advantages to other in-pile methods since it does not perturb the reactor system and consequently does not "excite" any sort of harmonic modes. As a by-product and a consistency check, the eff parameter was obtained without the need of the Diven factor and power normalization, and it is in excellent agreement with independent measurements. The theory/experiment comparison shows that for the abundances the JENDL3.3 presents the best performance, while for the decay constants the revised version of ENDF/B-VI.8 shows the best agreement. The best performance for the eff determination is obtained with JENDL3.3. In contrast, ENDF/B-VI.8 and its revised version performed at Los Alamos National Laboratory overestimate eff by as much as 4%. The eff results of this work totally support the proposal by Sakurai and Okajima to reduce the thermal delayed neutron yield of 235U.