Cepstral analysis in the speakers recognition systems

The present paper addresses issues related to the speaker recognition system (ASR - Automatic Speakers Recognition). In its primary form, a speech signal is characterized by a high redundancy, so it is necessary to extract the specific features of the signal that would allow to efficiently describing the properties thereof that are important from the viewpoint of speaker recognition. Therefore, parameterization of the signal in the process of recognition is extremely important. The authors have attempted to select the optimal (most discriminating) set of parameters describing the signal by using a homomorphic processing method. The study has primarily focused on assessing applicability of the cepstral analysis in speech recognition systems based on the acquired digitized voice samples.