The wavelet theory is a powerful tool for processing and compressing time-series or images. To summarize, the wavelet transform consists to project a signal on an orthornormal basis of functions. The sets of functions is chosen in order to provide a sparse representation of the initial time-series. In the first part of this article, we use wavelets for smoothing mortality curves. Mortality curves are projected into the space of Daubechies wavelets and we use a chi-square test for reducing the dimension of the curves by thresholding the smallest wavelet coefficients. In the second part, we study the evolution of wavelet coefficients for the Belgian population from 1950 till today. Next we forecast the wavelet coefficients in order to predict the evolution of mortality and compare with alternative methods, like the Lee-Carter model.