Cepstrum and LPC based statistical approach to explore singers' rendition on a set of songs of Tagore

The paper presents a systematic procedure using cepstrum and linear predictive coding (LPC) based statistical approach for exploring the singers rendition for a set of song of poet Rabindra Nath Tagore sung by different renowned singers. The medium category (Mudara) of songs started with vowel ‘aa’ has been incorporated in the data set of song. Index of perceived vocal qualities like flow phonation, depth, vowel quality and timbre have been estimated numerically based on cepstral and LPC coefficients. The value of probability of significance calculated by ANOVA has been considered for the comparative study of flow phonation, depth, vowel quality and timbre of the singers and the results show the significant difference among the singers of same song of Tagore.