Effect of GCI perturbation on speech quality in Indian languages

Abstract

Harmonic plus noise model (HNM) divides the speech spectrum into two bands: harmonic and noise. One is modeled with harmonics of the fundamental and the other is simulated using random noise. As HNM analysis/synthesis is performed at each glottal closure instant (GCI), errors in estimation of GCIs affect the quality of the synthesized speech. The objective of this paper is to investigate the effect of the perturbation in GCIs on the synthesized speech quality. Childers and Hu's algorithm is used for GCIs calculations. Investigations show that the speech quality is very sensitive to positions of the GCIs. A small perturbation with maximum of 4 % of the local pitch period considerably degrades the synthesized speech. Perturbations above 8 % severely affect quality of the output speech.