The Lee-Carter (LC) model is an elegant and powerful methodology to forecast mortality based on a log-bilinear form for the force of mortality. While it has been widely accepted and adopted, the model has some limitations that hinder its performance. Here, we propose a novel extension of the LC model that addresses the drawback of the fixed rate of mortality improvement ß_x. In particular, we introduce a three-component Lee-Carter (3C-LC) model based on the decomposition of mortality into infant, adult and senescent mortality. The 3C-LC model two major advantages. First, forecast mortality patterns do not depend on a single fixed ß_x, but rather on the combination of three component-specific schedules. Second, mortality forecasts can be derived from the projection of only three time-varying parameters. We illustrate and compare the 3C-LC methodology with the standard LC model by estimating and forecasting mortality patterns for Swiss males; analyses will be extended to a wider set of countries in the Human Mortality Database.