EN
In this paper a new construction of functional principal components (FPCA) is proposed, based on principal components for vector data. A kernel version of FPCA is also presented. The quality of the two described methods was tested on 20 different data sets.