The objective of this dissertation is to present a new measure of core inflation to brazilian inflation in order to verify the current trend of prices. Initially two core inflation are proposed. The first one is calculated from traditional methods found in the literature with some changes proposed in order to achieve some desirable core inflation characteristics. The second method of estimation is Dynamic Conditional Score Models (DCS), where inflation is decomposed into unobservables components and the core is defined as the trend component. A comparison is made between the proposed cores measures and that ones currently available in Brazil. The conclusion is that the core inflation disclosed today do not contribute to the understanding of the inflation trend. The two proposed core inflation measures are better suited for this purpose. The core by DCS model is further indicated because its estimation is not affected by lagged information as it happens for the first core measure. In addition, some economic analyzes suggest that the use of the core by DCS model can help the Central Bank in the conduct of monetary policy, since this shows indicative of the future development of the accumulated inflation in the year (measure used to verify the inflation target in Brazil). The two proposed methodologies can easily be recalculated, since implementation codes are freely available to any interested user, and it can also be replicated to other inflation indicators.