Smooth Cepstrum Calculation Using Modified Bartlett Hanning Window

Cepstrum is an algorithm for analyzing the speech signals in frequency domain. This is conventional method of fundamental peak picking i.e. fundamental frequency or pitch. For a speech signal it is necessary to identify the fundamental frequency correctly in order to have robust system for speaker identification and verification. Using this approach two algorithms has been proposed using Hamming window and Modified Bartlett Hanning Window. These algorithms involve the filtering called liftering after converting the speech sample to Cepstrum domain (Quefrency domain). The utility of using filtering in cepstrum domain has been analyzed as window functions play vital role in deciding the fundamental peak of the speech sample. Gross Pitch Estimation Error has been calculated with smooth cepstrum and smooth cepstrum with autocorrelation in order to calculate false acceptance and rejection rate for two speakers respectively.