The thesis is focused on getting acquainted with methods of adapting speech signals in spectrum domain and with formants. Fist part is dealing with definition of formant, the relation between Czech language and formants, performing formant features and describing the use of formants in practice. This part includes preprocessing of signals before analysis on its own. This work contains a list of the most common methods for counting speech signals spectrum focused on estimation of spectrum by linear prediction. Practical part contains a verification of the most common methods for specifying formants, which are linear prediction coding and Fourier transformation. Next part is dealing with the description of methods used in created program. It ...