Power spectral density of speech in Serbian language

Power spectral density represents one of important characteristics of speech. The estimation of its value is done using various approaches and algorithms. In this paper, the power spectral density of speech in Serbian language is calculated by applying severel methods. For that purpose,. the speech base is formed in a such way that 8 speakers read the chosen phonetically balanced text. The effects of several parameters in corresponding methods for power spectral density estimation are analyzed. The results obtained by using different duration of speech samples, parameters of recording, date when the recording is done and speakers are analyzed and mutually compared. The mean values of power spectral density of speech in Serbian language are compared to the corresponding values of speech in other languages.