Three main speech recognition methods as well as features commonly used in speech recognition were discussed in the work. Main attention was paid to formant feature extraction methods. For one of the methods feature extraction in C++ was developed. It is based on linear prediction coding (LPC) parameters and calculates complex roots of characteristic polynomial. These roots are used for formant frequencies determination. Spectral envelope of speech signal and formant frequencies are displayed and stored in text files. Program user guide is presented in the work description. Investigation results of formant features for various speech sounds are presented in the appendices.